From 97705d3071654cebd6c4f0492432372b8840977b Mon Sep 17 00:00:00 2001 From: Alan Modra Date: Wed, 20 Nov 2013 14:18:33 +1030 Subject: [PATCH 3/3] Tidy powerpc*-linux support This separates the 32-bit sysv/linux/bsd code from the 64-bit linux code, and makes it possible to link code compiled with different options to those used to compile libffi. For example, a -mlong-double-128 libffi can be used with -mlong-double-64 code. Tested powerpc-linux, powerpc64-linux, powerpc64le-linux, and the hacky powerpc-freebsd simulated setup I used earlier. I've tried a few combinations of ABI variations; -mlong-double-64/-mlong-double-128 and -msvr4-struct-return/-maix-struct-return pass the testuite with flying colours. Testing -msoft-float ran into lack of soft-float multilibs on my system, but I looked at some testcases under gdb and they seemed to do the right thing. I also tested backward compatibility on powerpc-linux and powerpc64-linux by building libfff-3.0.13, then running the testsuite so that tests were built linked against the 3.0.13 libffi.so but ran with the new libffi. * Makefile.am (EXTRA_DIST): Add new src/powerpc files. (nodist_libffi_la_SOURCES ): Likewise. * configure.ac (HAVE_LONG_DOUBLE_VARIANT): Define for powerpc. * include/ffi.h.in (ffi_prep_types): Declare. * src/prep_cif.c (ffi_prep_cif_core): Call ffi_prep_types. * src/types.c (FFI_NONCONST_TYPEDEF): Define and use for HAVE_LONG_DOUBLE_VARIANT. * src/powerpc/ffi_powerpc.h: New file. * src/powerpc/ffi.c: Split into.. * src/powerpc/ffi_sysv.c: ..new file, and.. * src/powerpc/ffi_linux64.c: ..new file, rewriting parts. * src/powerpc/ffitarget.h (enum ffi_abi): Rewrite powerpc ABI selection as bits controlling features. * src/powerpc/linux64.S: For consistency, use POWERPC64 rather than __powerpc64__. * src/powerpc/linux64_closure.S: Likewise. * src/powerpc/ppc_closure.S: Likewise. Move .note.FNU-stack inside guard. * src/powerpc/sysv.S: Likewise. * configure: Regenerate. * fficonfig.h.in: Regenerate. * Makefile.in: Regenerate. Index: libffi-3.0.13/Makefile.am =================================================================== --- libffi-3.0.13.orig/Makefile.am +++ libffi-3.0.13/Makefile.am @@ -20,7 +20,9 @@ EXTRA_DIST = LICENSE ChangeLog.v1 Change src/m32r/ffi.c src/m32r/sysv.S src/m32r/ffitarget.h \ src/m68k/ffi.c src/m68k/sysv.S src/m68k/ffitarget.h \ src/microblaze/ffi.c src/microblaze/sysv.S \ - src/microblaze/ffitarget.h src/powerpc/ffi.c \ + src/microblaze/ffitarget.h \ + src/powerpc/ffi.c src/powerpc/ffi_powerpc.h \ + src/powerpc/ffi_sysv.c src/powerpc/ffi_linux64.c \ src/powerpc/sysv.S src/powerpc/linux64.S \ src/powerpc/linux64_closure.S src/powerpc/ppc_closure.S \ src/powerpc/asm.h src/powerpc/aix.S src/powerpc/darwin.S \ @@ -156,7 +158,7 @@ if MICROBLAZE nodist_libffi_la_SOURCES += src/microblaze/ffi.c src/microblaze/sysv.S endif if POWERPC -nodist_libffi_la_SOURCES += src/powerpc/ffi.c src/powerpc/sysv.S src/powerpc/ppc_closure.S src/powerpc/linux64.S src/powerpc/linux64_closure.S +nodist_libffi_la_SOURCES += src/powerpc/ffi.c src/powerpc/ffi_sysv.c src/powerpc/ffi_linux64.c src/powerpc/sysv.S src/powerpc/ppc_closure.S src/powerpc/linux64.S src/powerpc/linux64_closure.S endif if POWERPC_AIX nodist_libffi_la_SOURCES += src/powerpc/ffi_darwin.c src/powerpc/aix.S src/powerpc/aix_closure.S @@ -165,7 +167,7 @@ if POWERPC_DARWIN nodist_libffi_la_SOURCES += src/powerpc/ffi_darwin.c src/powerpc/darwin.S src/powerpc/darwin_closure.S endif if POWERPC_FREEBSD -nodist_libffi_la_SOURCES += src/powerpc/ffi.c src/powerpc/sysv.S src/powerpc/ppc_closure.S +nodist_libffi_la_SOURCES += src/powerpc/ffi.c src/powerpc/ffi_sysv.c src/powerpc/sysv.S src/powerpc/ppc_closure.S endif if AARCH64 nodist_libffi_la_SOURCES += src/aarch64/sysv.S src/aarch64/ffi.c Index: libffi-3.0.13/configure.ac =================================================================== --- libffi-3.0.13.orig/configure.ac +++ libffi-3.0.13/configure.ac @@ -66,6 +66,7 @@ dnl The -no-testsuite modules omit the t AM_CONDITIONAL(TESTSUBDIR, test -d $srcdir/testsuite) TARGETDIR="unknown" +HAVE_LONG_DOUBLE_VARIANT=0 case "$host" in aarch64*-*-*) TARGET=AARCH64; TARGETDIR=aarch64 @@ -214,6 +215,7 @@ case "$host" in powerpc*-*-linux* | powerpc-*-sysv*) TARGET=POWERPC; TARGETDIR=powerpc + HAVE_LONG_DOUBLE_VARIANT=1 ;; powerpc-*-amigaos*) TARGET=POWERPC; TARGETDIR=powerpc @@ -229,6 +231,7 @@ case "$host" in ;; powerpc-*-freebsd* | powerpc-*-openbsd*) TARGET=POWERPC_FREEBSD; TARGETDIR=powerpc + HAVE_LONG_DOUBLE_VARIANT=1 ;; powerpc64-*-freebsd*) TARGET=POWERPC; TARGETDIR=powerpc @@ -313,14 +316,20 @@ AC_CHECK_SIZEOF(long double) # Also AC_SUBST this variable for ffi.h. if test -z "$HAVE_LONG_DOUBLE"; then HAVE_LONG_DOUBLE=0 - if test $ac_cv_sizeof_double != $ac_cv_sizeof_long_double; then - if test $ac_cv_sizeof_long_double != 0; then + if test $ac_cv_sizeof_long_double != 0; then + if test $HAVE_LONG_DOUBLE_VARIANT != 0; then + AC_DEFINE(HAVE_LONG_DOUBLE_VARIANT, 1, [Define if you support more than one size of the long double type]) HAVE_LONG_DOUBLE=1 - AC_DEFINE(HAVE_LONG_DOUBLE, 1, [Define if you have the long double type and it is bigger than a double]) + else + if test $ac_cv_sizeof_double != $ac_cv_sizeof_long_double; then + HAVE_LONG_DOUBLE=1 + AC_DEFINE(HAVE_LONG_DOUBLE, 1, [Define if you have the long double type and it is bigger than a double]) + fi fi fi fi AC_SUBST(HAVE_LONG_DOUBLE) +AC_SUBST(HAVE_LONG_DOUBLE_VARIANT) AC_C_BIGENDIAN Index: libffi-3.0.13/include/ffi.h.in =================================================================== --- libffi-3.0.13.orig/include/ffi.h.in +++ libffi-3.0.13/include/ffi.h.in @@ -221,6 +221,11 @@ typedef struct { #endif } ffi_cif; +#if HAVE_LONG_DOUBLE_VARIANT +/* Used to adjust size/alignment of ffi types. */ +void ffi_prep_types (ffi_abi abi); +# endif + /* Used internally, but overridden by some architectures */ ffi_status ffi_prep_cif_core(ffi_cif *cif, ffi_abi abi, Index: libffi-3.0.13/src/powerpc/ffi.c =================================================================== --- libffi-3.0.13.orig/src/powerpc/ffi.c +++ libffi-3.0.13/src/powerpc/ffi.c @@ -1,5 +1,6 @@ /* ----------------------------------------------------------------------- - ffi.c - Copyright (C) 2011 Anthony Green + ffi.c - Copyright (C) 2013 IBM + Copyright (C) 2011 Anthony Green Copyright (C) 2011 Kyle Moffett Copyright (C) 2008 Red Hat, Inc Copyright (C) 2007, 2008 Free Software Foundation, Inc @@ -27,1112 +28,60 @@ OTHER DEALINGS IN THE SOFTWARE. ----------------------------------------------------------------------- */ -#include -#include - -#include -#include - - -extern void ffi_closure_SYSV (void); -extern void FFI_HIDDEN ffi_closure_LINUX64 (void); - -enum { - /* The assembly depends on these exact flags. */ - FLAG_RETURNS_SMST = 1 << (31-31), /* Used for FFI_SYSV small structs. */ - FLAG_RETURNS_NOTHING = 1 << (31-30), /* These go in cr7 */ -#ifndef __NO_FPRS__ - FLAG_RETURNS_FP = 1 << (31-29), -#endif - FLAG_RETURNS_64BITS = 1 << (31-28), - - FLAG_RETURNS_128BITS = 1 << (31-27), /* cr6 */ - - FLAG_ARG_NEEDS_COPY = 1 << (31- 7), - FLAG_ARG_NEEDS_PSAVE = FLAG_ARG_NEEDS_COPY, /* Used by ELFv2 */ -#ifndef __NO_FPRS__ - FLAG_FP_ARGUMENTS = 1 << (31- 6), /* cr1.eq; specified by ABI */ -#endif - FLAG_4_GPR_ARGUMENTS = 1 << (31- 5), - FLAG_RETVAL_REFERENCE = 1 << (31- 4) -}; - -/* About the SYSV ABI. */ -#define ASM_NEEDS_REGISTERS 4 -#define NUM_GPR_ARG_REGISTERS 8 -#ifndef __NO_FPRS__ -# define NUM_FPR_ARG_REGISTERS 8 -#endif - -/* ffi_prep_args_SYSV is called by the assembly routine once stack space - has been allocated for the function's arguments. - - The stack layout we want looks like this: - - | Return address from ffi_call_SYSV 4bytes | higher addresses - |--------------------------------------------| - | Previous backchain pointer 4 | stack pointer here - |--------------------------------------------|<+ <<< on entry to - | Saved r28-r31 4*4 | | ffi_call_SYSV - |--------------------------------------------| | - | GPR registers r3-r10 8*4 | | ffi_call_SYSV - |--------------------------------------------| | - | FPR registers f1-f8 (optional) 8*8 | | - |--------------------------------------------| | stack | - | Space for copied structures | | grows | - |--------------------------------------------| | down V - | Parameters that didn't fit in registers | | - |--------------------------------------------| | lower addresses - | Space for callee's LR 4 | | - |--------------------------------------------| | stack pointer here - | Current backchain pointer 4 |-/ during - |--------------------------------------------| <<< ffi_call_SYSV - -*/ - -void -ffi_prep_args_SYSV (extended_cif *ecif, unsigned *const stack) -{ - const unsigned bytes = ecif->cif->bytes; - const unsigned flags = ecif->cif->flags; - - typedef union { - char *c; - unsigned *u; - long long *ll; - float *f; - double *d; - } valp; - - /* 'stacktop' points at the previous backchain pointer. */ - valp stacktop; - - /* 'gpr_base' points at the space for gpr3, and grows upwards as - we use GPR registers. */ - valp gpr_base; - int intarg_count; - -#ifndef __NO_FPRS__ - /* 'fpr_base' points at the space for fpr1, and grows upwards as - we use FPR registers. */ - valp fpr_base; - int fparg_count; -#endif - - /* 'copy_space' grows down as we put structures in it. It should - stay 16-byte aligned. */ - valp copy_space; - - /* 'next_arg' grows up as we put parameters in it. */ - valp next_arg; - - int i; - ffi_type **ptr; -#ifndef __NO_FPRS__ - double double_tmp; -#endif - union { - void **v; - char **c; - signed char **sc; - unsigned char **uc; - signed short **ss; - unsigned short **us; - unsigned int **ui; - long long **ll; - float **f; - double **d; - } p_argv; - size_t struct_copy_size; - unsigned gprvalue; - - stacktop.c = (char *) stack + bytes; - gpr_base.u = stacktop.u - ASM_NEEDS_REGISTERS - NUM_GPR_ARG_REGISTERS; - intarg_count = 0; -#ifndef __NO_FPRS__ - fpr_base.d = gpr_base.d - NUM_FPR_ARG_REGISTERS; - fparg_count = 0; - copy_space.c = ((flags & FLAG_FP_ARGUMENTS) ? fpr_base.c : gpr_base.c); -#else - copy_space.c = gpr_base.c; -#endif - next_arg.u = stack + 2; - - /* Check that everything starts aligned properly. */ - FFI_ASSERT (((unsigned long) (char *) stack & 0xF) == 0); - FFI_ASSERT (((unsigned long) copy_space.c & 0xF) == 0); - FFI_ASSERT (((unsigned long) stacktop.c & 0xF) == 0); - FFI_ASSERT ((bytes & 0xF) == 0); - FFI_ASSERT (copy_space.c >= next_arg.c); - - /* Deal with return values that are actually pass-by-reference. */ - if (flags & FLAG_RETVAL_REFERENCE) - { - *gpr_base.u++ = (unsigned long) (char *) ecif->rvalue; - intarg_count++; - } - - /* Now for the arguments. */ - p_argv.v = ecif->avalue; - for (ptr = ecif->cif->arg_types, i = ecif->cif->nargs; - i > 0; - i--, ptr++, p_argv.v++) - { - unsigned short typenum = (*ptr)->type; - - /* We may need to handle some values depending on ABI */ - if (ecif->cif->abi == FFI_LINUX_SOFT_FLOAT) { - if (typenum == FFI_TYPE_FLOAT) - typenum = FFI_TYPE_UINT32; - if (typenum == FFI_TYPE_DOUBLE) - typenum = FFI_TYPE_UINT64; - if (typenum == FFI_TYPE_LONGDOUBLE) - typenum = FFI_TYPE_UINT128; - } else if (ecif->cif->abi != FFI_LINUX) { -#if FFI_TYPE_LONGDOUBLE != FFI_TYPE_DOUBLE - if (typenum == FFI_TYPE_LONGDOUBLE) - typenum = FFI_TYPE_STRUCT; -#endif - } - - /* Now test the translated value */ - switch (typenum) { -#ifndef __NO_FPRS__ - case FFI_TYPE_FLOAT: - /* With FFI_LINUX_SOFT_FLOAT floats are handled like UINT32. */ - double_tmp = **p_argv.f; - if (fparg_count >= NUM_FPR_ARG_REGISTERS) - { - *next_arg.f = (float) double_tmp; - next_arg.u += 1; - intarg_count++; - } - else - *fpr_base.d++ = double_tmp; - fparg_count++; - FFI_ASSERT (flags & FLAG_FP_ARGUMENTS); - break; - - case FFI_TYPE_DOUBLE: - /* With FFI_LINUX_SOFT_FLOAT doubles are handled like UINT64. */ - double_tmp = **p_argv.d; - - if (fparg_count >= NUM_FPR_ARG_REGISTERS) - { - if (intarg_count >= NUM_GPR_ARG_REGISTERS - && intarg_count % 2 != 0) - { - intarg_count++; - next_arg.u++; - } - *next_arg.d = double_tmp; - next_arg.u += 2; - } - else - *fpr_base.d++ = double_tmp; - fparg_count++; - FFI_ASSERT (flags & FLAG_FP_ARGUMENTS); - break; - -#if FFI_TYPE_LONGDOUBLE != FFI_TYPE_DOUBLE - case FFI_TYPE_LONGDOUBLE: - double_tmp = (*p_argv.d)[0]; - - if (fparg_count >= NUM_FPR_ARG_REGISTERS - 1) - { - if (intarg_count >= NUM_GPR_ARG_REGISTERS - && intarg_count % 2 != 0) - { - intarg_count++; - next_arg.u++; - } - *next_arg.d = double_tmp; - next_arg.u += 2; - double_tmp = (*p_argv.d)[1]; - *next_arg.d = double_tmp; - next_arg.u += 2; - } - else - { - *fpr_base.d++ = double_tmp; - double_tmp = (*p_argv.d)[1]; - *fpr_base.d++ = double_tmp; - } - - fparg_count += 2; - FFI_ASSERT (flags & FLAG_FP_ARGUMENTS); - break; -#endif -#endif /* have FPRs */ - - /* - * The soft float ABI for long doubles works like this, a long double - * is passed in four consecutive GPRs if available. A maximum of 2 - * long doubles can be passed in gprs. If we do not have 4 GPRs - * left, the long double is passed on the stack, 4-byte aligned. - */ - case FFI_TYPE_UINT128: { - unsigned int int_tmp = (*p_argv.ui)[0]; - unsigned int ii; - if (intarg_count >= NUM_GPR_ARG_REGISTERS - 3) { - if (intarg_count < NUM_GPR_ARG_REGISTERS) - intarg_count += NUM_GPR_ARG_REGISTERS - intarg_count; - *(next_arg.u++) = int_tmp; - for (ii = 1; ii < 4; ii++) { - int_tmp = (*p_argv.ui)[ii]; - *(next_arg.u++) = int_tmp; - } - } else { - *(gpr_base.u++) = int_tmp; - for (ii = 1; ii < 4; ii++) { - int_tmp = (*p_argv.ui)[ii]; - *(gpr_base.u++) = int_tmp; - } - } - intarg_count += 4; - break; - } - - case FFI_TYPE_UINT64: - case FFI_TYPE_SINT64: - if (intarg_count == NUM_GPR_ARG_REGISTERS-1) - intarg_count++; - if (intarg_count >= NUM_GPR_ARG_REGISTERS) - { - if (intarg_count % 2 != 0) - { - intarg_count++; - next_arg.u++; - } - *next_arg.ll = **p_argv.ll; - next_arg.u += 2; - } - else - { - /* whoops: abi states only certain register pairs - * can be used for passing long long int - * specifically (r3,r4), (r5,r6), (r7,r8), - * (r9,r10) and if next arg is long long but - * not correct starting register of pair then skip - * until the proper starting register - */ - if (intarg_count % 2 != 0) - { - intarg_count ++; - gpr_base.u++; - } - *gpr_base.ll++ = **p_argv.ll; - } - intarg_count += 2; - break; - - case FFI_TYPE_STRUCT: - struct_copy_size = ((*ptr)->size + 15) & ~0xF; - copy_space.c -= struct_copy_size; - memcpy (copy_space.c, *p_argv.c, (*ptr)->size); - - gprvalue = (unsigned long) copy_space.c; - - FFI_ASSERT (copy_space.c > next_arg.c); - FFI_ASSERT (flags & FLAG_ARG_NEEDS_COPY); - goto putgpr; - - case FFI_TYPE_UINT8: - gprvalue = **p_argv.uc; - goto putgpr; - case FFI_TYPE_SINT8: - gprvalue = **p_argv.sc; - goto putgpr; - case FFI_TYPE_UINT16: - gprvalue = **p_argv.us; - goto putgpr; - case FFI_TYPE_SINT16: - gprvalue = **p_argv.ss; - goto putgpr; - - case FFI_TYPE_INT: - case FFI_TYPE_UINT32: - case FFI_TYPE_SINT32: - case FFI_TYPE_POINTER: - - gprvalue = **p_argv.ui; - - putgpr: - if (intarg_count >= NUM_GPR_ARG_REGISTERS) - *next_arg.u++ = gprvalue; - else - *gpr_base.u++ = gprvalue; - intarg_count++; - break; - } - } - - /* Check that we didn't overrun the stack... */ - FFI_ASSERT (copy_space.c >= next_arg.c); - FFI_ASSERT (gpr_base.u <= stacktop.u - ASM_NEEDS_REGISTERS); - /* The assert below is testing that the number of integer arguments agrees - with the number found in ffi_prep_cif_machdep(). However, intarg_count - is incremented whenever we place an FP arg on the stack, so account for - that before our assert test. */ -#ifndef __NO_FPRS__ - if (fparg_count > NUM_FPR_ARG_REGISTERS) - intarg_count -= fparg_count - NUM_FPR_ARG_REGISTERS; - FFI_ASSERT (fpr_base.u - <= stacktop.u - ASM_NEEDS_REGISTERS - NUM_GPR_ARG_REGISTERS); -#endif - FFI_ASSERT (flags & FLAG_4_GPR_ARGUMENTS || intarg_count <= 4); -} - -/* About the LINUX64 ABI. */ -enum { - NUM_GPR_ARG_REGISTERS64 = 8, - NUM_FPR_ARG_REGISTERS64 = 13 -}; -enum { ASM_NEEDS_REGISTERS64 = 4 }; - -#if _CALL_ELF == 2 -static unsigned int -discover_homogeneous_aggregate (const ffi_type *t, unsigned int *elnum) -{ - switch (t->type) - { - case FFI_TYPE_FLOAT: - case FFI_TYPE_DOUBLE: - *elnum = 1; - return (int) t->type; - - case FFI_TYPE_STRUCT:; - { - unsigned int base_elt = 0, total_elnum = 0; - ffi_type **el = t->elements; - while (*el) - { - unsigned int el_elt, el_elnum = 0; - el_elt = discover_homogeneous_aggregate (*el, &el_elnum); - if (el_elt == 0 - || (base_elt && base_elt != el_elt)) - return 0; - base_elt = el_elt; - total_elnum += el_elnum; - if (total_elnum > 8) - return 0; - el++; - } - *elnum = total_elnum; - return base_elt; - } - - default: - return 0; - } -} -#endif - - -/* ffi_prep_args64 is called by the assembly routine once stack space - has been allocated for the function's arguments. - - The stack layout we want looks like this: - - | Ret addr from ffi_call_LINUX64 8bytes | higher addresses - |--------------------------------------------| - | CR save area 8bytes | - |--------------------------------------------| - | Previous backchain pointer 8 | stack pointer here - |--------------------------------------------|<+ <<< on entry to - | Saved r28-r31 4*8 | | ffi_call_LINUX64 - |--------------------------------------------| | - | GPR registers r3-r10 8*8 | | - |--------------------------------------------| | - | FPR registers f1-f13 (optional) 13*8 | | - |--------------------------------------------| | - | Parameter save area | | - |--------------------------------------------| | - | TOC save area 8 | | - |--------------------------------------------| | stack | - | Linker doubleword 8 | | grows | - |--------------------------------------------| | down V - | Compiler doubleword 8 | | - |--------------------------------------------| | lower addresses - | Space for callee's LR 8 | | - |--------------------------------------------| | - | CR save area 8 | | - |--------------------------------------------| | stack pointer here - | Current backchain pointer 8 |-/ during - |--------------------------------------------| <<< ffi_call_LINUX64 - -*/ +#include "ffi.h" +#include "ffi_common.h" +#include "ffi_powerpc.h" +#if HAVE_LONG_DOUBLE_VARIANT +/* Adjust ffi_type_longdouble. */ void FFI_HIDDEN -ffi_prep_args64 (extended_cif *ecif, unsigned long *const stack) +ffi_prep_types (ffi_abi abi) { - const unsigned long bytes = ecif->cif->bytes; - const unsigned long flags = ecif->cif->flags; - - typedef union { - char *c; - unsigned long *ul; - float *f; - double *d; - size_t p; - } valp; - - /* 'stacktop' points at the previous backchain pointer. */ - valp stacktop; - - /* 'next_arg' points at the space for gpr3, and grows upwards as - we use GPR registers, then continues at rest. */ - valp gpr_base; - valp gpr_end; - valp rest; - valp next_arg; - - /* 'fpr_base' points at the space for fpr3, and grows upwards as - we use FPR registers. */ - valp fpr_base; - unsigned int fparg_count; - - unsigned int i, words, nargs, nfixedargs; - ffi_type **ptr; - double double_tmp; - union { - void **v; - char **c; - signed char **sc; - unsigned char **uc; - signed short **ss; - unsigned short **us; - signed int **si; - unsigned int **ui; - unsigned long **ul; - float **f; - double **d; - } p_argv; - unsigned long gprvalue; -#ifdef __STRUCT_PARM_ALIGN__ - unsigned long align; -#endif - - stacktop.c = (char *) stack + bytes; - gpr_base.ul = stacktop.ul - ASM_NEEDS_REGISTERS64 - NUM_GPR_ARG_REGISTERS64; - gpr_end.ul = gpr_base.ul + NUM_GPR_ARG_REGISTERS64; -#if _CALL_ELF == 2 - rest.ul = stack + 4 + NUM_GPR_ARG_REGISTERS64; -#else - rest.ul = stack + 6 + NUM_GPR_ARG_REGISTERS64; -#endif - fpr_base.d = gpr_base.d - NUM_FPR_ARG_REGISTERS64; - fparg_count = 0; - next_arg.ul = gpr_base.ul; - - /* Check that everything starts aligned properly. */ - FFI_ASSERT (((unsigned long) (char *) stack & 0xF) == 0); - FFI_ASSERT (((unsigned long) stacktop.c & 0xF) == 0); - FFI_ASSERT ((bytes & 0xF) == 0); - - /* Deal with return values that are actually pass-by-reference. */ - if (flags & FLAG_RETVAL_REFERENCE) - *next_arg.ul++ = (unsigned long) (char *) ecif->rvalue; - - /* Now for the arguments. */ - p_argv.v = ecif->avalue; - nargs = ecif->cif->nargs; - nfixedargs = ecif->cif->nfixedargs; - for (ptr = ecif->cif->arg_types, i = 0; - i < nargs; - i++, ptr++, p_argv.v++) - { - unsigned int elt, elnum; - - switch ((*ptr)->type) - { - case FFI_TYPE_FLOAT: - double_tmp = **p_argv.f; - if (fparg_count < NUM_FPR_ARG_REGISTERS64 && i < nfixedargs) - *fpr_base.d++ = double_tmp; - else - *next_arg.f = (float) double_tmp; - if (++next_arg.ul == gpr_end.ul) - next_arg.ul = rest.ul; - fparg_count++; - FFI_ASSERT (flags & FLAG_FP_ARGUMENTS); - break; - - case FFI_TYPE_DOUBLE: - double_tmp = **p_argv.d; - if (fparg_count < NUM_FPR_ARG_REGISTERS64 && i < nfixedargs) - *fpr_base.d++ = double_tmp; - else - *next_arg.d = double_tmp; - if (++next_arg.ul == gpr_end.ul) - next_arg.ul = rest.ul; - fparg_count++; - FFI_ASSERT (flags & FLAG_FP_ARGUMENTS); - break; - -#if FFI_TYPE_LONGDOUBLE != FFI_TYPE_DOUBLE - case FFI_TYPE_LONGDOUBLE: - double_tmp = (*p_argv.d)[0]; - if (fparg_count < NUM_FPR_ARG_REGISTERS64 && i < nfixedargs) - *fpr_base.d++ = double_tmp; - else - *next_arg.d = double_tmp; - if (++next_arg.ul == gpr_end.ul) - next_arg.ul = rest.ul; - fparg_count++; - double_tmp = (*p_argv.d)[1]; - if (fparg_count < NUM_FPR_ARG_REGISTERS64 && i < nfixedargs) - *fpr_base.d++ = double_tmp; - else - *next_arg.d = double_tmp; - if (++next_arg.ul == gpr_end.ul) - next_arg.ul = rest.ul; - fparg_count++; - FFI_ASSERT (__LDBL_MANT_DIG__ == 106); - FFI_ASSERT (flags & FLAG_FP_ARGUMENTS); - break; -#endif - - case FFI_TYPE_STRUCT: -#ifdef __STRUCT_PARM_ALIGN__ - align = (*ptr)->alignment; - if (align > __STRUCT_PARM_ALIGN__) - align = __STRUCT_PARM_ALIGN__; - if (align > 1) - next_arg.p = ALIGN (next_arg.p, align); -#endif - elt = 0; -#if _CALL_ELF == 2 - elt = discover_homogeneous_aggregate (*ptr, &elnum); -#endif - if (elt) - { - union { - void *v; - float *f; - double *d; - } arg; - - arg.v = *p_argv.v; - if (elt == FFI_TYPE_FLOAT) - { - do - { - double_tmp = *arg.f++; - if (fparg_count < NUM_FPR_ARG_REGISTERS64 - && i < nfixedargs) - *fpr_base.d++ = double_tmp; - else - *next_arg.f = (float) double_tmp; - if (++next_arg.f == gpr_end.f) - next_arg.f = rest.f; - fparg_count++; - } - while (--elnum != 0); - if ((next_arg.p & 3) != 0) - { - if (++next_arg.f == gpr_end.f) - next_arg.f = rest.f; - } - } - else - do - { - double_tmp = *arg.d++; - if (fparg_count < NUM_FPR_ARG_REGISTERS64 && i < nfixedargs) - *fpr_base.d++ = double_tmp; - else - *next_arg.d = double_tmp; - if (++next_arg.d == gpr_end.d) - next_arg.d = rest.d; - fparg_count++; - } - while (--elnum != 0); - } - else - { - words = ((*ptr)->size + 7) / 8; - if (next_arg.ul >= gpr_base.ul && next_arg.ul + words > gpr_end.ul) - { - size_t first = gpr_end.c - next_arg.c; - memcpy (next_arg.c, *p_argv.c, first); - memcpy (rest.c, *p_argv.c + first, (*ptr)->size - first); - next_arg.c = rest.c + words * 8 - first; - } - else - { - char *where = next_arg.c; - -#ifndef __LITTLE_ENDIAN__ - /* Structures with size less than eight bytes are passed - left-padded. */ - if ((*ptr)->size < 8) - where += 8 - (*ptr)->size; -#endif - memcpy (where, *p_argv.c, (*ptr)->size); - next_arg.ul += words; - if (next_arg.ul == gpr_end.ul) - next_arg.ul = rest.ul; - } - } - break; - - case FFI_TYPE_UINT8: - gprvalue = **p_argv.uc; - goto putgpr; - case FFI_TYPE_SINT8: - gprvalue = **p_argv.sc; - goto putgpr; - case FFI_TYPE_UINT16: - gprvalue = **p_argv.us; - goto putgpr; - case FFI_TYPE_SINT16: - gprvalue = **p_argv.ss; - goto putgpr; - case FFI_TYPE_UINT32: - gprvalue = **p_argv.ui; - goto putgpr; - case FFI_TYPE_INT: - case FFI_TYPE_SINT32: - gprvalue = **p_argv.si; - goto putgpr; - - case FFI_TYPE_UINT64: - case FFI_TYPE_SINT64: - case FFI_TYPE_POINTER: - gprvalue = **p_argv.ul; - putgpr: - *next_arg.ul++ = gprvalue; - if (next_arg.ul == gpr_end.ul) - next_arg.ul = rest.ul; - break; - } - } - - FFI_ASSERT (flags & FLAG_4_GPR_ARGUMENTS - || (next_arg.ul >= gpr_base.ul - && next_arg.ul <= gpr_base.ul + 4)); +# if FFI_TYPE_LONGDOUBLE != FFI_TYPE_DOUBLE +# ifdef POWERPC64 + ffi_prep_types_linux64 (abi); +# else + ffi_prep_types_sysv (abi); +# endif +# endif } - - +#endif /* Perform machine dependent cif processing */ -static ffi_status -ffi_prep_cif_machdep_core (ffi_cif *cif) +ffi_status FFI_HIDDEN +ffi_prep_cif_machdep (ffi_cif *cif) { - /* All this is for the SYSV and LINUX64 ABI. */ - ffi_type **ptr; - unsigned bytes; - unsigned i, fparg_count = 0, intarg_count = 0; - unsigned flags = cif->flags; - unsigned struct_copy_size = 0; - unsigned type = cif->rtype->type; - unsigned size = cif->rtype->size; - - /* The machine-independent calculation of cif->bytes doesn't work - for us. Redo the calculation. */ - if (cif->abi != FFI_LINUX64) - { - /* Space for the frame pointer, callee's LR, and the asm's temp regs. */ - bytes = (2 + ASM_NEEDS_REGISTERS) * sizeof (int); - - /* Space for the GPR registers. */ - bytes += NUM_GPR_ARG_REGISTERS * sizeof (int); - } - else - { - /* 64-bit ABI. */ -#if _CALL_ELF == 2 - /* Space for backchain, CR, LR, TOC and the asm's temp regs. */ - bytes = (4 + ASM_NEEDS_REGISTERS64) * sizeof (long); - - /* Space for the general registers. */ - bytes += NUM_GPR_ARG_REGISTERS64 * sizeof (long); -#else - /* Space for backchain, CR, LR, cc/ld doubleword, TOC and the asm's temp - regs. */ - bytes = (6 + ASM_NEEDS_REGISTERS64) * sizeof (long); - - /* Space for the mandatory parm save area and general registers. */ - bytes += 2 * NUM_GPR_ARG_REGISTERS64 * sizeof (long); -#endif - } - - /* Return value handling. The rules for SYSV are as follows: - - 32-bit (or less) integer values are returned in gpr3; - - Structures of size <= 4 bytes also returned in gpr3; - - 64-bit integer values and structures between 5 and 8 bytes are returned - in gpr3 and gpr4; - - Single/double FP values are returned in fpr1; - - Larger structures are allocated space and a pointer is passed as - the first argument. - - long doubles (if not equivalent to double) are returned in - fpr1,fpr2 for Linux and as for large structs for SysV. - For LINUX64: - - integer values in gpr3; - - Structures/Unions by reference; - - Single/double FP values in fpr1, long double in fpr1,fpr2. - - soft-float float/doubles are treated as UINT32/UINT64 respectivley. - - soft-float long doubles are returned in gpr3-gpr6. */ - /* First translate for softfloat/nonlinux */ - if (cif->abi == FFI_LINUX_SOFT_FLOAT) - { - if (type == FFI_TYPE_FLOAT) - type = FFI_TYPE_UINT32; - if (type == FFI_TYPE_DOUBLE) - type = FFI_TYPE_UINT64; - if (type == FFI_TYPE_LONGDOUBLE) - type = FFI_TYPE_UINT128; - } - else if (cif->abi != FFI_LINUX - && cif->abi != FFI_LINUX64) - { -#if FFI_TYPE_LONGDOUBLE != FFI_TYPE_DOUBLE - if (type == FFI_TYPE_LONGDOUBLE) - type = FFI_TYPE_STRUCT; -#endif - } - - switch (type) - { -#ifndef __NO_FPRS__ -#if FFI_TYPE_LONGDOUBLE != FFI_TYPE_DOUBLE - case FFI_TYPE_LONGDOUBLE: - flags |= FLAG_RETURNS_128BITS; - /* Fall through. */ -#endif - case FFI_TYPE_DOUBLE: - flags |= FLAG_RETURNS_64BITS; - /* Fall through. */ - case FFI_TYPE_FLOAT: - flags |= FLAG_RETURNS_FP; - break; -#endif - - case FFI_TYPE_UINT128: - flags |= FLAG_RETURNS_128BITS; - /* Fall through. */ - case FFI_TYPE_UINT64: - case FFI_TYPE_SINT64: - flags |= FLAG_RETURNS_64BITS; - break; - - case FFI_TYPE_STRUCT: - /* - * The final SYSV ABI says that structures smaller or equal 8 bytes - * are returned in r3/r4. The FFI_GCC_SYSV ABI instead returns them - * in memory. - * - * NOTE: The assembly code can safely assume that it just needs to - * store both r3 and r4 into a 8-byte word-aligned buffer, as - * we allocate a temporary buffer in ffi_call() if this flag is - * set. - */ - if (cif->abi == FFI_SYSV && size <= 8) - { - flags |= FLAG_RETURNS_SMST; - break; - } -#if _CALL_ELF == 2 - if (cif->abi == FFI_LINUX64) - { - unsigned int elt, elnum; - elt = discover_homogeneous_aggregate (cif->rtype, &elnum); - if (elt) - { - if (elt == FFI_TYPE_DOUBLE) - flags |= FLAG_RETURNS_64BITS; - flags |= FLAG_RETURNS_FP | FLAG_RETURNS_SMST; - break; - } - if (size <= 16) - { - flags |= FLAG_RETURNS_SMST; - break; - } - } -#endif - intarg_count++; - flags |= FLAG_RETVAL_REFERENCE; - /* Fall through. */ - case FFI_TYPE_VOID: - flags |= FLAG_RETURNS_NOTHING; - break; - - default: - /* Returns 32-bit integer, or similar. Nothing to do here. */ - break; - } - - if (cif->abi != FFI_LINUX64) - /* The first NUM_GPR_ARG_REGISTERS words of integer arguments, and the - first NUM_FPR_ARG_REGISTERS fp arguments, go in registers; the rest - goes on the stack. Structures and long doubles (if not equivalent - to double) are passed as a pointer to a copy of the structure. - Stuff on the stack needs to keep proper alignment. */ - for (ptr = cif->arg_types, i = cif->nargs; i > 0; i--, ptr++) - { - unsigned short typenum = (*ptr)->type; - - /* We may need to handle some values depending on ABI */ - if (cif->abi == FFI_LINUX_SOFT_FLOAT) { - if (typenum == FFI_TYPE_FLOAT) - typenum = FFI_TYPE_UINT32; - if (typenum == FFI_TYPE_DOUBLE) - typenum = FFI_TYPE_UINT64; - if (typenum == FFI_TYPE_LONGDOUBLE) - typenum = FFI_TYPE_UINT128; - } else if (cif->abi != FFI_LINUX && cif->abi != FFI_LINUX64) { -#if FFI_TYPE_LONGDOUBLE != FFI_TYPE_DOUBLE - if (typenum == FFI_TYPE_LONGDOUBLE) - typenum = FFI_TYPE_STRUCT; -#endif - } - - switch (typenum) { -#ifndef __NO_FPRS__ - case FFI_TYPE_FLOAT: - fparg_count++; - /* floating singles are not 8-aligned on stack */ - break; - -#if FFI_TYPE_LONGDOUBLE != FFI_TYPE_DOUBLE - case FFI_TYPE_LONGDOUBLE: - fparg_count++; - /* Fall thru */ -#endif - case FFI_TYPE_DOUBLE: - fparg_count++; - /* If this FP arg is going on the stack, it must be - 8-byte-aligned. */ - if (fparg_count > NUM_FPR_ARG_REGISTERS - && intarg_count >= NUM_GPR_ARG_REGISTERS - && intarg_count % 2 != 0) - intarg_count++; - break; -#endif - case FFI_TYPE_UINT128: - /* - * A long double in FFI_LINUX_SOFT_FLOAT can use only a set - * of four consecutive gprs. If we do not have enough, we - * have to adjust the intarg_count value. - */ - if (intarg_count >= NUM_GPR_ARG_REGISTERS - 3 - && intarg_count < NUM_GPR_ARG_REGISTERS) - intarg_count = NUM_GPR_ARG_REGISTERS; - intarg_count += 4; - break; - - case FFI_TYPE_UINT64: - case FFI_TYPE_SINT64: - /* 'long long' arguments are passed as two words, but - either both words must fit in registers or both go - on the stack. If they go on the stack, they must - be 8-byte-aligned. - - Also, only certain register pairs can be used for - passing long long int -- specifically (r3,r4), (r5,r6), - (r7,r8), (r9,r10). - */ - if (intarg_count == NUM_GPR_ARG_REGISTERS-1 - || intarg_count % 2 != 0) - intarg_count++; - intarg_count += 2; - break; - - case FFI_TYPE_STRUCT: - /* We must allocate space for a copy of these to enforce - pass-by-value. Pad the space up to a multiple of 16 - bytes (the maximum alignment required for anything under - the SYSV ABI). */ - struct_copy_size += ((*ptr)->size + 15) & ~0xF; - /* Fall through (allocate space for the pointer). */ - - case FFI_TYPE_POINTER: - case FFI_TYPE_INT: - case FFI_TYPE_UINT32: - case FFI_TYPE_SINT32: - case FFI_TYPE_UINT16: - case FFI_TYPE_SINT16: - case FFI_TYPE_UINT8: - case FFI_TYPE_SINT8: - /* Everything else is passed as a 4-byte word in a GPR, either - the object itself or a pointer to it. */ - intarg_count++; - break; - default: - FFI_ASSERT (0); - } - } - else - for (ptr = cif->arg_types, i = cif->nargs; i > 0; i--, ptr++) - { - unsigned int elt, elnum; -#ifdef __STRUCT_PARM_ALIGN__ - unsigned int align; -#endif - - switch ((*ptr)->type) - { -#if FFI_TYPE_LONGDOUBLE != FFI_TYPE_DOUBLE - case FFI_TYPE_LONGDOUBLE: - fparg_count += 2; - intarg_count += 2; - if (fparg_count > NUM_FPR_ARG_REGISTERS64) - flags |= FLAG_ARG_NEEDS_PSAVE; - break; -#endif - case FFI_TYPE_FLOAT: - case FFI_TYPE_DOUBLE: - fparg_count++; - intarg_count++; - if (fparg_count > NUM_FPR_ARG_REGISTERS64) - flags |= FLAG_ARG_NEEDS_PSAVE; - break; - - case FFI_TYPE_STRUCT: -#ifdef __STRUCT_PARM_ALIGN__ - align = (*ptr)->alignment; - if (align > __STRUCT_PARM_ALIGN__) - align = __STRUCT_PARM_ALIGN__; - align = align / 8; - if (align > 1) - intarg_count = ALIGN (intarg_count, align); -#endif - intarg_count += ((*ptr)->size + 7) / 8; - elt = 0; -#if _CALL_ELF == 2 - elt = discover_homogeneous_aggregate (*ptr, &elnum); -#endif - if (elt) - { - fparg_count += elnum; - if (fparg_count > NUM_FPR_ARG_REGISTERS64) - flags |= FLAG_ARG_NEEDS_PSAVE; - } - else - { - if (intarg_count > NUM_GPR_ARG_REGISTERS64) - flags |= FLAG_ARG_NEEDS_PSAVE; - } - break; - - case FFI_TYPE_POINTER: - case FFI_TYPE_UINT64: - case FFI_TYPE_SINT64: - case FFI_TYPE_INT: - case FFI_TYPE_UINT32: - case FFI_TYPE_SINT32: - case FFI_TYPE_UINT16: - case FFI_TYPE_SINT16: - case FFI_TYPE_UINT8: - case FFI_TYPE_SINT8: - /* Everything else is passed as a 8-byte word in a GPR, either - the object itself or a pointer to it. */ - intarg_count++; - if (intarg_count > NUM_GPR_ARG_REGISTERS64) - flags |= FLAG_ARG_NEEDS_PSAVE; - break; - default: - FFI_ASSERT (0); - } - } - -#ifndef __NO_FPRS__ - if (fparg_count != 0) - flags |= FLAG_FP_ARGUMENTS; -#endif - if (intarg_count > 4) - flags |= FLAG_4_GPR_ARGUMENTS; - if (struct_copy_size != 0) - flags |= FLAG_ARG_NEEDS_COPY; - - if (cif->abi != FFI_LINUX64) - { -#ifndef __NO_FPRS__ - /* Space for the FPR registers, if needed. */ - if (fparg_count != 0) - bytes += NUM_FPR_ARG_REGISTERS * sizeof (double); -#endif - - /* Stack space. */ - if (intarg_count > NUM_GPR_ARG_REGISTERS) - bytes += (intarg_count - NUM_GPR_ARG_REGISTERS) * sizeof (int); -#ifndef __NO_FPRS__ - if (fparg_count > NUM_FPR_ARG_REGISTERS) - bytes += (fparg_count - NUM_FPR_ARG_REGISTERS) * sizeof (double); -#endif - } - else - { -#ifndef __NO_FPRS__ - /* Space for the FPR registers, if needed. */ - if (fparg_count != 0) - bytes += NUM_FPR_ARG_REGISTERS64 * sizeof (double); -#endif - - /* Stack space. */ -#if _CALL_ELF == 2 - if ((flags & FLAG_ARG_NEEDS_PSAVE) != 0) - bytes += intarg_count * sizeof (long); +#ifdef POWERPC64 + return ffi_prep_cif_linux64 (cif); #else - if (intarg_count > NUM_GPR_ARG_REGISTERS64) - bytes += (intarg_count - NUM_GPR_ARG_REGISTERS64) * sizeof (long); + return ffi_prep_cif_sysv (cif); #endif - } - - /* The stack space allocated needs to be a multiple of 16 bytes. */ - bytes = (bytes + 15) & ~0xF; - - /* Add in the space for the copied structures. */ - bytes += struct_copy_size; - - cif->flags = flags; - cif->bytes = bytes; - - return FFI_OK; -} - -ffi_status -ffi_prep_cif_machdep (ffi_cif *cif) -{ - cif->nfixedargs = cif->nargs; - return ffi_prep_cif_machdep_core (cif); } -ffi_status +ffi_status FFI_HIDDEN ffi_prep_cif_machdep_var (ffi_cif *cif, - unsigned int nfixedargs, + unsigned int nfixedargs MAYBE_UNUSED, unsigned int ntotalargs MAYBE_UNUSED) { - cif->nfixedargs = nfixedargs; -#if _CALL_ELF == 2 - if (cif->abi == FFI_LINUX64) - cif->flags |= FLAG_ARG_NEEDS_PSAVE; +#ifdef POWERPC64 + return ffi_prep_cif_linux64_var (cif, nfixedargs, ntotalargs); +#else + return ffi_prep_cif_sysv (cif); #endif - return ffi_prep_cif_machdep_core (cif); } -extern void ffi_call_SYSV(extended_cif *, unsigned, unsigned, unsigned *, - void (*fn)(void)); -extern void FFI_HIDDEN ffi_call_LINUX64(extended_cif *, unsigned long, - unsigned long, unsigned long *, - void (*fn)(void)); - void ffi_call(ffi_cif *cif, void (*fn)(void), void *rvalue, void **avalue) { - /* - * The final SYSV ABI says that structures smaller or equal 8 bytes - * are returned in r3/r4. The FFI_GCC_SYSV ABI instead returns them - * in memory. - * - * We bounce-buffer SYSV small struct return values so that sysv.S - * can write r3 and r4 to memory without worrying about struct size. - * - * For ELFv2 ABI, use a bounce buffer for homogeneous structs too, - * for similar reasons. - */ + /* The final SYSV ABI says that structures smaller or equal 8 bytes + are returned in r3/r4. A draft ABI used by linux instead returns + them in memory. + + We bounce-buffer SYSV small struct return values so that sysv.S + can write r3 and r4 to memory without worrying about struct size. + + For ELFv2 ABI, use a bounce buffer for homogeneous structs too, + for similar reasons. */ unsigned long smst_buffer[8]; extended_cif ecif; @@ -1147,26 +96,11 @@ ffi_call(ffi_cif *cif, void (*fn)(void), else if (!rvalue && cif->rtype->type == FFI_TYPE_STRUCT) ecif.rvalue = alloca (cif->rtype->size); - switch (cif->abi) - { -#ifndef POWERPC64 -# ifndef __NO_FPRS__ - case FFI_SYSV: - case FFI_GCC_SYSV: - case FFI_LINUX: -# endif - case FFI_LINUX_SOFT_FLOAT: - ffi_call_SYSV (&ecif, -cif->bytes, cif->flags, ecif.rvalue, fn); - break; +#ifdef POWERPC64 + ffi_call_LINUX64 (&ecif, -(long) cif->bytes, cif->flags, ecif.rvalue, fn); #else - case FFI_LINUX64: - ffi_call_LINUX64 (&ecif, -(long) cif->bytes, cif->flags, ecif.rvalue, fn); - break; + ffi_call_SYSV (&ecif, -cif->bytes, cif->flags, ecif.rvalue, fn); #endif - default: - FFI_ASSERT (0); - break; - } /* Check for a bounce-buffered return value */ if (rvalue && ecif.rvalue == smst_buffer) @@ -1175,36 +109,23 @@ ffi_call(ffi_cif *cif, void (*fn)(void), #ifndef __LITTLE_ENDIAN__ /* The SYSV ABI returns a structure of up to 4 bytes in size left-padded in r3. */ - if (cif->abi == FFI_SYSV && rsize <= 4) +# ifndef POWERPC64 + if (rsize <= 4) memcpy (rvalue, (char *) smst_buffer + 4 - rsize, rsize); - /* The SYSV ABI returns a structure of up to 8 bytes in size - left-padded in r3/r4, and the ELFv2 ABI similarly returns a - structure of up to 8 bytes in size left-padded in r3. */ - else if (rsize <= 8) - memcpy (rvalue, (char *) smst_buffer + 8 - rsize, rsize); else +# endif + /* The SYSV ABI returns a structure of up to 8 bytes in size + left-padded in r3/r4, and the ELFv2 ABI similarly returns a + structure of up to 8 bytes in size left-padded in r3. */ + if (rsize <= 8) + memcpy (rvalue, (char *) smst_buffer + 8 - rsize, rsize); + else #endif - memcpy (rvalue, smst_buffer, rsize); + memcpy (rvalue, smst_buffer, rsize); } } -#if !defined POWERPC64 || _CALL_ELF == 2 -#define MIN_CACHE_LINE_SIZE 8 - -static void -flush_icache (char *wraddr, char *xaddr, int size) -{ - int i; - for (i = 0; i < size; i += MIN_CACHE_LINE_SIZE) - __asm__ volatile ("icbi 0,%0;" "dcbf 0,%1;" - : : "r" (xaddr + i), "r" (wraddr + i) : "memory"); - __asm__ volatile ("icbi 0,%0;" "dcbf 0,%1;" "sync;" "isync;" - : : "r"(xaddr + size - 1), "r"(wraddr + size - 1) - : "memory"); -} -#endif - ffi_status ffi_prep_closure_loc (ffi_closure *closure, ffi_cif *cif, @@ -1213,593 +134,8 @@ ffi_prep_closure_loc (ffi_closure *closu void *codeloc) { #ifdef POWERPC64 -# if _CALL_ELF == 2 - unsigned int *tramp = (unsigned int *) &closure->tramp[0]; - - if (cif->abi != FFI_LINUX64) - return FFI_BAD_ABI; - - tramp[0] = 0xe96c0018; /* 0: ld 11,2f-0b(12) */ - tramp[1] = 0xe98c0010; /* ld 12,1f-0b(12) */ - tramp[2] = 0x7d8903a6; /* mtctr 12 */ - tramp[3] = 0x4e800420; /* bctr */ - /* 1: .quad function_addr */ - /* 2: .quad context */ - *(void **) &tramp[4] = (void *) ffi_closure_LINUX64; - *(void **) &tramp[6] = codeloc; - flush_icache ((char *)tramp, (char *)codeloc, FFI_TRAMPOLINE_SIZE); -# else - void **tramp = (void **) &closure->tramp[0]; - - if (cif->abi != FFI_LINUX64) - return FFI_BAD_ABI; - /* Copy function address and TOC from ffi_closure_LINUX64. */ - memcpy (tramp, (char *) ffi_closure_LINUX64, 16); - tramp[2] = codeloc; -# endif + return ffi_prep_closure_loc_linux64 (closure, cif, fun, user_data, codeloc); #else - unsigned int *tramp; - - if (! (cif->abi == FFI_GCC_SYSV - || cif->abi == FFI_SYSV - || cif->abi == FFI_LINUX - || cif->abi == FFI_LINUX_SOFT_FLOAT)) - return FFI_BAD_ABI; - - tramp = (unsigned int *) &closure->tramp[0]; - tramp[0] = 0x7c0802a6; /* mflr r0 */ - tramp[1] = 0x4800000d; /* bl 10 */ - tramp[4] = 0x7d6802a6; /* mflr r11 */ - tramp[5] = 0x7c0803a6; /* mtlr r0 */ - tramp[6] = 0x800b0000; /* lwz r0,0(r11) */ - tramp[7] = 0x816b0004; /* lwz r11,4(r11) */ - tramp[8] = 0x7c0903a6; /* mtctr r0 */ - tramp[9] = 0x4e800420; /* bctr */ - *(void **) &tramp[2] = (void *) ffi_closure_SYSV; /* function */ - *(void **) &tramp[3] = codeloc; /* context */ - - /* Flush the icache. */ - flush_icache ((char *)tramp, (char *)codeloc, FFI_TRAMPOLINE_SIZE); -#endif - - closure->cif = cif; - closure->fun = fun; - closure->user_data = user_data; - - return FFI_OK; -} - -typedef union -{ - float f; - double d; -} ffi_dblfl; - -int ffi_closure_helper_SYSV (ffi_closure *, void *, unsigned long *, - ffi_dblfl *, unsigned long *); - -/* Basically the trampoline invokes ffi_closure_SYSV, and on - * entry, r11 holds the address of the closure. - * After storing the registers that could possibly contain - * parameters to be passed into the stack frame and setting - * up space for a return value, ffi_closure_SYSV invokes the - * following helper function to do most of the work - */ - -int -ffi_closure_helper_SYSV (ffi_closure *closure, void *rvalue, - unsigned long *pgr, ffi_dblfl *pfr, - unsigned long *pst) -{ - /* rvalue is the pointer to space for return value in closure assembly */ - /* pgr is the pointer to where r3-r10 are stored in ffi_closure_SYSV */ - /* pfr is the pointer to where f1-f8 are stored in ffi_closure_SYSV */ - /* pst is the pointer to outgoing parameter stack in original caller */ - - void ** avalue; - ffi_type ** arg_types; - long i, avn; -#ifndef __NO_FPRS__ - long nf = 0; /* number of floating registers already used */ -#endif - long ng = 0; /* number of general registers already used */ - - ffi_cif *cif = closure->cif; - unsigned size = cif->rtype->size; - unsigned short rtypenum = cif->rtype->type; - - avalue = alloca (cif->nargs * sizeof (void *)); - - /* First translate for softfloat/nonlinux */ - if (cif->abi == FFI_LINUX_SOFT_FLOAT) { - if (rtypenum == FFI_TYPE_FLOAT) - rtypenum = FFI_TYPE_UINT32; - if (rtypenum == FFI_TYPE_DOUBLE) - rtypenum = FFI_TYPE_UINT64; - if (rtypenum == FFI_TYPE_LONGDOUBLE) - rtypenum = FFI_TYPE_UINT128; - } else if (cif->abi != FFI_LINUX && cif->abi != FFI_LINUX64) { -#if FFI_TYPE_LONGDOUBLE != FFI_TYPE_DOUBLE - if (rtypenum == FFI_TYPE_LONGDOUBLE) - rtypenum = FFI_TYPE_STRUCT; -#endif - } - - - /* Copy the caller's structure return value address so that the closure - returns the data directly to the caller. - For FFI_SYSV the result is passed in r3/r4 if the struct size is less - or equal 8 bytes. */ - if (rtypenum == FFI_TYPE_STRUCT && ((cif->abi != FFI_SYSV) || (size > 8))) { - rvalue = (void *) *pgr; - ng++; - pgr++; - } - - i = 0; - avn = cif->nargs; - arg_types = cif->arg_types; - - /* Grab the addresses of the arguments from the stack frame. */ - while (i < avn) { - unsigned short typenum = arg_types[i]->type; - - /* We may need to handle some values depending on ABI */ - if (cif->abi == FFI_LINUX_SOFT_FLOAT) { - if (typenum == FFI_TYPE_FLOAT) - typenum = FFI_TYPE_UINT32; - if (typenum == FFI_TYPE_DOUBLE) - typenum = FFI_TYPE_UINT64; - if (typenum == FFI_TYPE_LONGDOUBLE) - typenum = FFI_TYPE_UINT128; - } else if (cif->abi != FFI_LINUX && cif->abi != FFI_LINUX64) { -#if FFI_TYPE_LONGDOUBLE != FFI_TYPE_DOUBLE - if (typenum == FFI_TYPE_LONGDOUBLE) - typenum = FFI_TYPE_STRUCT; + return ffi_prep_closure_loc_sysv (closure, cif, fun, user_data, codeloc); #endif - } - - switch (typenum) { -#ifndef __NO_FPRS__ - case FFI_TYPE_FLOAT: - /* unfortunately float values are stored as doubles - * in the ffi_closure_SYSV code (since we don't check - * the type in that routine). - */ - - /* there are 8 64bit floating point registers */ - - if (nf < 8) - { - double temp = pfr->d; - pfr->f = (float) temp; - avalue[i] = pfr; - nf++; - pfr++; - } - else - { - /* FIXME? here we are really changing the values - * stored in the original calling routines outgoing - * parameter stack. This is probably a really - * naughty thing to do but... - */ - avalue[i] = pst; - pst += 1; - } - break; - - case FFI_TYPE_DOUBLE: - /* On the outgoing stack all values are aligned to 8 */ - /* there are 8 64bit floating point registers */ - - if (nf < 8) - { - avalue[i] = pfr; - nf++; - pfr++; - } - else - { - if (((long) pst) & 4) - pst++; - avalue[i] = pst; - pst += 2; - } - break; - -#if FFI_TYPE_LONGDOUBLE != FFI_TYPE_DOUBLE - case FFI_TYPE_LONGDOUBLE: - if (nf < 7) - { - avalue[i] = pfr; - pfr += 2; - nf += 2; - } - else - { - if (((long) pst) & 4) - pst++; - avalue[i] = pst; - pst += 4; - nf = 8; - } - break; -#endif -#endif /* have FPRS */ - - case FFI_TYPE_UINT128: - /* - * Test if for the whole long double, 4 gprs are available. - * otherwise the stuff ends up on the stack. - */ - if (ng < 5) { - avalue[i] = pgr; - pgr += 4; - ng += 4; - } else { - avalue[i] = pst; - pst += 4; - ng = 8+4; - } - break; - - case FFI_TYPE_SINT8: - case FFI_TYPE_UINT8: -#ifndef __LITTLE_ENDIAN__ - /* there are 8 gpr registers used to pass values */ - if (ng < 8) - { - avalue[i] = (char *) pgr + 3; - ng++; - pgr++; - } - else - { - avalue[i] = (char *) pst + 3; - pst++; - } - break; -#endif - - case FFI_TYPE_SINT16: - case FFI_TYPE_UINT16: -#ifndef __LITTLE_ENDIAN__ - /* there are 8 gpr registers used to pass values */ - if (ng < 8) - { - avalue[i] = (char *) pgr + 2; - ng++; - pgr++; - } - else - { - avalue[i] = (char *) pst + 2; - pst++; - } - break; -#endif - - case FFI_TYPE_SINT32: - case FFI_TYPE_UINT32: - case FFI_TYPE_POINTER: - /* there are 8 gpr registers used to pass values */ - if (ng < 8) - { - avalue[i] = pgr; - ng++; - pgr++; - } - else - { - avalue[i] = pst; - pst++; - } - break; - - case FFI_TYPE_STRUCT: - /* Structs are passed by reference. The address will appear in a - gpr if it is one of the first 8 arguments. */ - if (ng < 8) - { - avalue[i] = (void *) *pgr; - ng++; - pgr++; - } - else - { - avalue[i] = (void *) *pst; - pst++; - } - break; - - case FFI_TYPE_SINT64: - case FFI_TYPE_UINT64: - /* passing long long ints are complex, they must - * be passed in suitable register pairs such as - * (r3,r4) or (r5,r6) or (r6,r7), or (r7,r8) or (r9,r10) - * and if the entire pair aren't available then the outgoing - * parameter stack is used for both but an alignment of 8 - * must will be kept. So we must either look in pgr - * or pst to find the correct address for this type - * of parameter. - */ - if (ng < 7) - { - if (ng & 0x01) - { - /* skip r4, r6, r8 as starting points */ - ng++; - pgr++; - } - avalue[i] = pgr; - ng += 2; - pgr += 2; - } - else - { - if (((long) pst) & 4) - pst++; - avalue[i] = pst; - pst += 2; - ng = 8; - } - break; - - default: - FFI_ASSERT (0); - } - - i++; - } - - - (closure->fun) (cif, rvalue, avalue, closure->user_data); - - /* Tell ffi_closure_SYSV how to perform return type promotions. - Because the FFI_SYSV ABI returns the structures <= 8 bytes in r3/r4 - we have to tell ffi_closure_SYSV how to treat them. We combine the base - type FFI_SYSV_TYPE_SMALL_STRUCT - 1 with the size of the struct. - So a one byte struct gets the return type 16. Return type 1 to 15 are - already used and we never have a struct with size zero. That is the reason - for the subtraction of 1. See the comment in ffitarget.h about ordering. - */ - if (cif->abi == FFI_SYSV && rtypenum == FFI_TYPE_STRUCT && size <= 8) - return (FFI_SYSV_TYPE_SMALL_STRUCT - 1) + size; - return rtypenum; -} - -int FFI_HIDDEN ffi_closure_helper_LINUX64 (ffi_closure *, void *, - unsigned long *, ffi_dblfl *); - -int FFI_HIDDEN -ffi_closure_helper_LINUX64 (ffi_closure *closure, void *rvalue, - unsigned long *pst, ffi_dblfl *pfr) -{ - /* rvalue is the pointer to space for return value in closure assembly */ - /* pst is the pointer to parameter save area - (r3-r10 are stored into its first 8 slots by ffi_closure_LINUX64) */ - /* pfr is the pointer to where f1-f13 are stored in ffi_closure_LINUX64 */ - - void **avalue; - ffi_type **arg_types; - unsigned long i, avn, nfixedargs; - ffi_cif *cif; - ffi_dblfl *end_pfr = pfr + NUM_FPR_ARG_REGISTERS64; -#ifdef __STRUCT_PARM_ALIGN__ - unsigned long align; -#endif - - cif = closure->cif; - avalue = alloca (cif->nargs * sizeof (void *)); - - /* Copy the caller's structure return value address so that the - closure returns the data directly to the caller. */ - if (cif->rtype->type == FFI_TYPE_STRUCT - && (cif->flags & FLAG_RETURNS_SMST) == 0) - { - rvalue = (void *) *pst; - pst++; - } - - i = 0; - avn = cif->nargs; - nfixedargs = cif->nfixedargs; - arg_types = cif->arg_types; - - /* Grab the addresses of the arguments from the stack frame. */ - while (i < avn) - { - unsigned int elt, elnum; - - switch (arg_types[i]->type) - { - case FFI_TYPE_SINT8: - case FFI_TYPE_UINT8: -#ifndef __LITTLE_ENDIAN__ - avalue[i] = (char *) pst + 7; - pst++; - break; -#endif - - case FFI_TYPE_SINT16: - case FFI_TYPE_UINT16: -#ifndef __LITTLE_ENDIAN__ - avalue[i] = (char *) pst + 6; - pst++; - break; -#endif - - case FFI_TYPE_SINT32: - case FFI_TYPE_UINT32: -#ifndef __LITTLE_ENDIAN__ - avalue[i] = (char *) pst + 4; - pst++; - break; -#endif - - case FFI_TYPE_SINT64: - case FFI_TYPE_UINT64: - case FFI_TYPE_POINTER: - avalue[i] = pst; - pst++; - break; - - case FFI_TYPE_STRUCT: -#ifdef __STRUCT_PARM_ALIGN__ - align = arg_types[i]->alignment; - if (align > __STRUCT_PARM_ALIGN__) - align = __STRUCT_PARM_ALIGN__; - if (align > 1) - pst = (unsigned long *) ALIGN ((size_t) pst, align); -#endif - elt = 0; -#if _CALL_ELF == 2 - elt = discover_homogeneous_aggregate (arg_types[i], &elnum); -#endif - if (elt) - { - union { - void *v; - unsigned long *ul; - float *f; - double *d; - size_t p; - } to, from; - - /* Repackage the aggregate from its parts. The - aggregate size is not greater than the space taken by - the registers so store back to the register/parameter - save arrays. */ - if (pfr + elnum <= end_pfr) - to.v = pfr; - else - to.v = pst; - - avalue[i] = to.v; - from.ul = pst; - if (elt == FFI_TYPE_FLOAT) - { - do - { - if (pfr < end_pfr && i < nfixedargs) - { - *to.f = (float) pfr->d; - pfr++; - } - else - *to.f = *from.f; - to.f++; - from.f++; - } - while (--elnum != 0); - } - else - { - do - { - if (pfr < end_pfr && i < nfixedargs) - { - *to.d = pfr->d; - pfr++; - } - else - *to.d = *from.d; - to.d++; - from.d++; - } - while (--elnum != 0); - } - } - else - { -#ifndef __LITTLE_ENDIAN__ - /* Structures with size less than eight bytes are passed - left-padded. */ - if (arg_types[i]->size < 8) - avalue[i] = (char *) pst + 8 - arg_types[i]->size; - else -#endif - avalue[i] = pst; - } - pst += (arg_types[i]->size + 7) / 8; - break; - - case FFI_TYPE_FLOAT: - /* unfortunately float values are stored as doubles - * in the ffi_closure_LINUX64 code (since we don't check - * the type in that routine). - */ - - /* there are 13 64bit floating point registers */ - - if (pfr < end_pfr && i < nfixedargs) - { - double temp = pfr->d; - pfr->f = (float) temp; - avalue[i] = pfr; - pfr++; - } - else - avalue[i] = pst; - pst++; - break; - - case FFI_TYPE_DOUBLE: - /* On the outgoing stack all values are aligned to 8 */ - /* there are 13 64bit floating point registers */ - - if (pfr < end_pfr && i < nfixedargs) - { - avalue[i] = pfr; - pfr++; - } - else - avalue[i] = pst; - pst++; - break; - -#if FFI_TYPE_LONGDOUBLE != FFI_TYPE_DOUBLE - case FFI_TYPE_LONGDOUBLE: - if (pfr + 1 < end_pfr && i + 1 < nfixedargs) - { - avalue[i] = pfr; - pfr += 2; - } - else - { - if (pfr < end_pfr && i < nfixedargs) - { - /* Passed partly in f13 and partly on the stack. - Move it all to the stack. */ - *pst = *(unsigned long *) pfr; - pfr++; - } - avalue[i] = pst; - } - pst += 2; - break; -#endif - - default: - FFI_ASSERT (0); - } - - i++; - } - - - (closure->fun) (cif, rvalue, avalue, closure->user_data); - - /* Tell ffi_closure_LINUX64 how to perform return type promotions. */ - if ((cif->flags & FLAG_RETURNS_SMST) != 0) - { - if ((cif->flags & FLAG_RETURNS_FP) == 0) - return FFI_V2_TYPE_SMALL_STRUCT + cif->rtype->size - 1; - else if ((cif->flags & FLAG_RETURNS_64BITS) != 0) - return FFI_V2_TYPE_DOUBLE_HOMOG; - else - return FFI_V2_TYPE_FLOAT_HOMOG; - } - return cif->rtype->type; } Index: libffi-3.0.13/src/powerpc/ffi_linux64.c =================================================================== --- /dev/null +++ libffi-3.0.13/src/powerpc/ffi_linux64.c @@ -0,0 +1,942 @@ +/* ----------------------------------------------------------------------- + ffi_linux64.c - Copyright (C) 2013 IBM + Copyright (C) 2011 Anthony Green + Copyright (C) 2011 Kyle Moffett + Copyright (C) 2008 Red Hat, Inc + Copyright (C) 2007, 2008 Free Software Foundation, Inc + Copyright (c) 1998 Geoffrey Keating + + PowerPC Foreign Function Interface + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + ``Software''), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice shall be included + in all copies or substantial portions of the Software. + + THE SOFTWARE IS PROVIDED ``AS IS'', WITHOUT WARRANTY OF ANY KIND, EXPRESS + OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY CLAIM, DAMAGES OR + OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + OTHER DEALINGS IN THE SOFTWARE. + ----------------------------------------------------------------------- */ + +#include "ffi.h" + +#ifdef POWERPC64 +#include "ffi_common.h" +#include "ffi_powerpc.h" + + +/* About the LINUX64 ABI. */ +enum { + NUM_GPR_ARG_REGISTERS64 = 8, + NUM_FPR_ARG_REGISTERS64 = 13 +}; +enum { ASM_NEEDS_REGISTERS64 = 4 }; + + +#if HAVE_LONG_DOUBLE_VARIANT && FFI_TYPE_LONGDOUBLE != FFI_TYPE_DOUBLE +/* Adjust size of ffi_type_longdouble. */ +void FFI_HIDDEN +ffi_prep_types_linux64 (ffi_abi abi) +{ + if ((abi & (FFI_LINUX | FFI_LINUX_LONG_DOUBLE_128)) == FFI_LINUX) + { + ffi_type_longdouble.size = 8; + ffi_type_longdouble.alignment = 8; + } + else + { + ffi_type_longdouble.size = 16; + ffi_type_longdouble.alignment = 16; + } +} +#endif + + +#if _CALL_ELF == 2 +static unsigned int +discover_homogeneous_aggregate (const ffi_type *t, unsigned int *elnum) +{ + switch (t->type) + { + case FFI_TYPE_FLOAT: + case FFI_TYPE_DOUBLE: + *elnum = 1; + return (int) t->type; + + case FFI_TYPE_STRUCT:; + { + unsigned int base_elt = 0, total_elnum = 0; + ffi_type **el = t->elements; + while (*el) + { + unsigned int el_elt, el_elnum = 0; + el_elt = discover_homogeneous_aggregate (*el, &el_elnum); + if (el_elt == 0 + || (base_elt && base_elt != el_elt)) + return 0; + base_elt = el_elt; + total_elnum += el_elnum; + if (total_elnum > 8) + return 0; + el++; + } + *elnum = total_elnum; + return base_elt; + } + + default: + return 0; + } +} +#endif + + +/* Perform machine dependent cif processing */ +static ffi_status +ffi_prep_cif_linux64_core (ffi_cif *cif) +{ + ffi_type **ptr; + unsigned bytes; + unsigned i, fparg_count = 0, intarg_count = 0; + unsigned flags = cif->flags; +#if _CALL_ELF == 2 + unsigned int elt, elnum; +#endif + +#if FFI_TYPE_LONGDOUBLE == FFI_TYPE_DOUBLE + /* If compiled without long double support.. */ + if ((cif->abi & FFI_LINUX_LONG_DOUBLE_128) != 0) + return FFI_BAD_ABI; +#endif + + /* The machine-independent calculation of cif->bytes doesn't work + for us. Redo the calculation. */ +#if _CALL_ELF == 2 + /* Space for backchain, CR, LR, TOC and the asm's temp regs. */ + bytes = (4 + ASM_NEEDS_REGISTERS64) * sizeof (long); + + /* Space for the general registers. */ + bytes += NUM_GPR_ARG_REGISTERS64 * sizeof (long); +#else + /* Space for backchain, CR, LR, cc/ld doubleword, TOC and the asm's temp + regs. */ + bytes = (6 + ASM_NEEDS_REGISTERS64) * sizeof (long); + + /* Space for the mandatory parm save area and general registers. */ + bytes += 2 * NUM_GPR_ARG_REGISTERS64 * sizeof (long); +#endif + + /* Return value handling. */ + switch (cif->rtype->type) + { +#if FFI_TYPE_LONGDOUBLE != FFI_TYPE_DOUBLE + case FFI_TYPE_LONGDOUBLE: + if ((cif->abi & FFI_LINUX_LONG_DOUBLE_128) != 0) + flags |= FLAG_RETURNS_128BITS; + /* Fall through. */ +#endif + case FFI_TYPE_DOUBLE: + flags |= FLAG_RETURNS_64BITS; + /* Fall through. */ + case FFI_TYPE_FLOAT: + flags |= FLAG_RETURNS_FP; + break; + + case FFI_TYPE_UINT128: + flags |= FLAG_RETURNS_128BITS; + /* Fall through. */ + case FFI_TYPE_UINT64: + case FFI_TYPE_SINT64: + flags |= FLAG_RETURNS_64BITS; + break; + + case FFI_TYPE_STRUCT: +#if _CALL_ELF == 2 + elt = discover_homogeneous_aggregate (cif->rtype, &elnum); + if (elt) + { + if (elt == FFI_TYPE_DOUBLE) + flags |= FLAG_RETURNS_64BITS; + flags |= FLAG_RETURNS_FP | FLAG_RETURNS_SMST; + break; + } + if (cif->rtype->size <= 16) + { + flags |= FLAG_RETURNS_SMST; + break; + } +#endif + intarg_count++; + flags |= FLAG_RETVAL_REFERENCE; + /* Fall through. */ + case FFI_TYPE_VOID: + flags |= FLAG_RETURNS_NOTHING; + break; + + default: + /* Returns 32-bit integer, or similar. Nothing to do here. */ + break; + } + + for (ptr = cif->arg_types, i = cif->nargs; i > 0; i--, ptr++) + { + unsigned int align; + + switch ((*ptr)->type) + { +#if FFI_TYPE_LONGDOUBLE != FFI_TYPE_DOUBLE + case FFI_TYPE_LONGDOUBLE: + if ((cif->abi & FFI_LINUX_LONG_DOUBLE_128) != 0) + { + fparg_count++; + intarg_count++; + } + /* Fall through. */ +#endif + case FFI_TYPE_DOUBLE: + case FFI_TYPE_FLOAT: + fparg_count++; + intarg_count++; + if (fparg_count > NUM_FPR_ARG_REGISTERS64) + flags |= FLAG_ARG_NEEDS_PSAVE; + break; + + case FFI_TYPE_STRUCT: + if ((cif->abi & FFI_LINUX_STRUCT_ALIGN) != 0) + { + align = (*ptr)->alignment; + if (align > 16) + align = 16; + align = align / 8; + if (align > 1) + intarg_count = ALIGN (intarg_count, align); + } + intarg_count += ((*ptr)->size + 7) / 8; +#if _CALL_ELF == 2 + elt = discover_homogeneous_aggregate (*ptr, &elnum); + if (elt) + { + fparg_count += elnum; + if (fparg_count > NUM_FPR_ARG_REGISTERS64) + flags |= FLAG_ARG_NEEDS_PSAVE; + } + else +#endif + { + if (intarg_count > NUM_GPR_ARG_REGISTERS64) + flags |= FLAG_ARG_NEEDS_PSAVE; + } + break; + + case FFI_TYPE_POINTER: + case FFI_TYPE_UINT64: + case FFI_TYPE_SINT64: + case FFI_TYPE_INT: + case FFI_TYPE_UINT32: + case FFI_TYPE_SINT32: + case FFI_TYPE_UINT16: + case FFI_TYPE_SINT16: + case FFI_TYPE_UINT8: + case FFI_TYPE_SINT8: + /* Everything else is passed as a 8-byte word in a GPR, either + the object itself or a pointer to it. */ + intarg_count++; + if (intarg_count > NUM_GPR_ARG_REGISTERS64) + flags |= FLAG_ARG_NEEDS_PSAVE; + break; + default: + FFI_ASSERT (0); + } + } + + if (fparg_count != 0) + flags |= FLAG_FP_ARGUMENTS; + if (intarg_count > 4) + flags |= FLAG_4_GPR_ARGUMENTS; + + /* Space for the FPR registers, if needed. */ + if (fparg_count != 0) + bytes += NUM_FPR_ARG_REGISTERS64 * sizeof (double); + + /* Stack space. */ +#if _CALL_ELF == 2 + if ((flags & FLAG_ARG_NEEDS_PSAVE) != 0) + bytes += intarg_count * sizeof (long); +#else + if (intarg_count > NUM_GPR_ARG_REGISTERS64) + bytes += (intarg_count - NUM_GPR_ARG_REGISTERS64) * sizeof (long); +#endif + + /* The stack space allocated needs to be a multiple of 16 bytes. */ + bytes = (bytes + 15) & ~0xF; + + cif->flags = flags; + cif->bytes = bytes; + + return FFI_OK; +} + +ffi_status FFI_HIDDEN +ffi_prep_cif_linux64 (ffi_cif *cif) +{ + if ((cif->abi & FFI_LINUX) != 0) + cif->nfixedargs = cif->nargs; +#if _CALL_ELF != 2 + else if (cif->abi == FFI_COMPAT_LINUX64) + { + /* This call is from old code. Don't touch cif->nfixedargs + since old code will be using a smaller cif. */ + cif->flags |= FLAG_COMPAT; + /* Translate to new abi value. */ + cif->abi = FFI_LINUX | FFI_LINUX_LONG_DOUBLE_128; + } +#endif + else + return FFI_BAD_ABI; + return ffi_prep_cif_linux64_core (cif); +} + +ffi_status FFI_HIDDEN +ffi_prep_cif_linux64_var (ffi_cif *cif, + unsigned int nfixedargs, + unsigned int ntotalargs MAYBE_UNUSED) +{ + if ((cif->abi & FFI_LINUX) != 0) + cif->nfixedargs = nfixedargs; +#if _CALL_ELF != 2 + else if (cif->abi == FFI_COMPAT_LINUX64) + { + /* This call is from old code. Don't touch cif->nfixedargs + since old code will be using a smaller cif. */ + cif->flags |= FLAG_COMPAT; + /* Translate to new abi value. */ + cif->abi = FFI_LINUX | FFI_LINUX_LONG_DOUBLE_128; + } +#endif + else + return FFI_BAD_ABI; +#if _CALL_ELF == 2 + cif->flags |= FLAG_ARG_NEEDS_PSAVE; +#endif + return ffi_prep_cif_linux64_core (cif); +} + + +/* ffi_prep_args64 is called by the assembly routine once stack space + has been allocated for the function's arguments. + + The stack layout we want looks like this: + + | Ret addr from ffi_call_LINUX64 8bytes | higher addresses + |--------------------------------------------| + | CR save area 8bytes | + |--------------------------------------------| + | Previous backchain pointer 8 | stack pointer here + |--------------------------------------------|<+ <<< on entry to + | Saved r28-r31 4*8 | | ffi_call_LINUX64 + |--------------------------------------------| | + | GPR registers r3-r10 8*8 | | + |--------------------------------------------| | + | FPR registers f1-f13 (optional) 13*8 | | + |--------------------------------------------| | + | Parameter save area | | + |--------------------------------------------| | + | TOC save area 8 | | + |--------------------------------------------| | stack | + | Linker doubleword 8 | | grows | + |--------------------------------------------| | down V + | Compiler doubleword 8 | | + |--------------------------------------------| | lower addresses + | Space for callee's LR 8 | | + |--------------------------------------------| | + | CR save area 8 | | + |--------------------------------------------| | stack pointer here + | Current backchain pointer 8 |-/ during + |--------------------------------------------| <<< ffi_call_LINUX64 + +*/ + +void FFI_HIDDEN +ffi_prep_args64 (extended_cif *ecif, unsigned long *const stack) +{ + const unsigned long bytes = ecif->cif->bytes; + const unsigned long flags = ecif->cif->flags; + + typedef union + { + char *c; + unsigned long *ul; + float *f; + double *d; + size_t p; + } valp; + + /* 'stacktop' points at the previous backchain pointer. */ + valp stacktop; + + /* 'next_arg' points at the space for gpr3, and grows upwards as + we use GPR registers, then continues at rest. */ + valp gpr_base; + valp gpr_end; + valp rest; + valp next_arg; + + /* 'fpr_base' points at the space for fpr3, and grows upwards as + we use FPR registers. */ + valp fpr_base; + unsigned int fparg_count; + + unsigned int i, words, nargs, nfixedargs; + ffi_type **ptr; + double double_tmp; + union + { + void **v; + char **c; + signed char **sc; + unsigned char **uc; + signed short **ss; + unsigned short **us; + signed int **si; + unsigned int **ui; + unsigned long **ul; + float **f; + double **d; + } p_argv; + unsigned long gprvalue; + unsigned long align; + + stacktop.c = (char *) stack + bytes; + gpr_base.ul = stacktop.ul - ASM_NEEDS_REGISTERS64 - NUM_GPR_ARG_REGISTERS64; + gpr_end.ul = gpr_base.ul + NUM_GPR_ARG_REGISTERS64; +#if _CALL_ELF == 2 + rest.ul = stack + 4 + NUM_GPR_ARG_REGISTERS64; +#else + rest.ul = stack + 6 + NUM_GPR_ARG_REGISTERS64; +#endif + fpr_base.d = gpr_base.d - NUM_FPR_ARG_REGISTERS64; + fparg_count = 0; + next_arg.ul = gpr_base.ul; + + /* Check that everything starts aligned properly. */ + FFI_ASSERT (((unsigned long) (char *) stack & 0xF) == 0); + FFI_ASSERT (((unsigned long) stacktop.c & 0xF) == 0); + FFI_ASSERT ((bytes & 0xF) == 0); + + /* Deal with return values that are actually pass-by-reference. */ + if (flags & FLAG_RETVAL_REFERENCE) + *next_arg.ul++ = (unsigned long) (char *) ecif->rvalue; + + /* Now for the arguments. */ + p_argv.v = ecif->avalue; + nargs = ecif->cif->nargs; +#if _CALL_ELF != 2 + nfixedargs = (unsigned) -1; + if ((flags & FLAG_COMPAT) == 0) +#endif + nfixedargs = ecif->cif->nfixedargs; + for (ptr = ecif->cif->arg_types, i = 0; + i < nargs; + i++, ptr++, p_argv.v++) + { +#if _CALL_ELF == 2 + unsigned int elt, elnum; +#endif + + switch ((*ptr)->type) + { +#if FFI_TYPE_LONGDOUBLE != FFI_TYPE_DOUBLE + case FFI_TYPE_LONGDOUBLE: + if ((ecif->cif->abi & FFI_LINUX_LONG_DOUBLE_128) != 0) + { + double_tmp = (*p_argv.d)[0]; + if (fparg_count < NUM_FPR_ARG_REGISTERS64 && i < nfixedargs) + { + *fpr_base.d++ = double_tmp; +# if _CALL_ELF != 2 + if ((flags & FLAG_COMPAT) != 0) + *next_arg.d = double_tmp; +# endif + } + else + *next_arg.d = double_tmp; + if (++next_arg.ul == gpr_end.ul) + next_arg.ul = rest.ul; + fparg_count++; + double_tmp = (*p_argv.d)[1]; + if (fparg_count < NUM_FPR_ARG_REGISTERS64 && i < nfixedargs) + { + *fpr_base.d++ = double_tmp; +# if _CALL_ELF != 2 + if ((flags & FLAG_COMPAT) != 0) + *next_arg.d = double_tmp; +# endif + } + else + *next_arg.d = double_tmp; + if (++next_arg.ul == gpr_end.ul) + next_arg.ul = rest.ul; + fparg_count++; + FFI_ASSERT (__LDBL_MANT_DIG__ == 106); + FFI_ASSERT (flags & FLAG_FP_ARGUMENTS); + break; + } + /* Fall through. */ +#endif + case FFI_TYPE_DOUBLE: + double_tmp = **p_argv.d; + if (fparg_count < NUM_FPR_ARG_REGISTERS64 && i < nfixedargs) + { + *fpr_base.d++ = double_tmp; +#if _CALL_ELF != 2 + if ((flags & FLAG_COMPAT) != 0) + *next_arg.d = double_tmp; +#endif + } + else + *next_arg.d = double_tmp; + if (++next_arg.ul == gpr_end.ul) + next_arg.ul = rest.ul; + fparg_count++; + FFI_ASSERT (flags & FLAG_FP_ARGUMENTS); + break; + + case FFI_TYPE_FLOAT: + double_tmp = **p_argv.f; + if (fparg_count < NUM_FPR_ARG_REGISTERS64 && i < nfixedargs) + { + *fpr_base.d++ = double_tmp; +#if _CALL_ELF != 2 + if ((flags & FLAG_COMPAT) != 0) + *next_arg.f = (float) double_tmp; +#endif + } + else + *next_arg.f = (float) double_tmp; + if (++next_arg.ul == gpr_end.ul) + next_arg.ul = rest.ul; + fparg_count++; + FFI_ASSERT (flags & FLAG_FP_ARGUMENTS); + break; + + case FFI_TYPE_STRUCT: + if ((ecif->cif->abi & FFI_LINUX_STRUCT_ALIGN) != 0) + { + align = (*ptr)->alignment; + if (align > 16) + align = 16; + if (align > 1) + next_arg.p = ALIGN (next_arg.p, align); + } +#if _CALL_ELF == 2 + elt = discover_homogeneous_aggregate (*ptr, &elnum); + if (elt) + { + union { + void *v; + float *f; + double *d; + } arg; + + arg.v = *p_argv.v; + if (elt == FFI_TYPE_FLOAT) + { + do + { + double_tmp = *arg.f++; + if (fparg_count < NUM_FPR_ARG_REGISTERS64 + && i < nfixedargs) + *fpr_base.d++ = double_tmp; + else + *next_arg.f = (float) double_tmp; + if (++next_arg.f == gpr_end.f) + next_arg.f = rest.f; + fparg_count++; + } + while (--elnum != 0); + if ((next_arg.p & 3) != 0) + { + if (++next_arg.f == gpr_end.f) + next_arg.f = rest.f; + } + } + else + do + { + double_tmp = *arg.d++; + if (fparg_count < NUM_FPR_ARG_REGISTERS64 && i < nfixedargs) + *fpr_base.d++ = double_tmp; + else + *next_arg.d = double_tmp; + if (++next_arg.d == gpr_end.d) + next_arg.d = rest.d; + fparg_count++; + } + while (--elnum != 0); + } + else +#endif + { + words = ((*ptr)->size + 7) / 8; + if (next_arg.ul >= gpr_base.ul && next_arg.ul + words > gpr_end.ul) + { + size_t first = gpr_end.c - next_arg.c; + memcpy (next_arg.c, *p_argv.c, first); + memcpy (rest.c, *p_argv.c + first, (*ptr)->size - first); + next_arg.c = rest.c + words * 8 - first; + } + else + { + char *where = next_arg.c; + +#ifndef __LITTLE_ENDIAN__ + /* Structures with size less than eight bytes are passed + left-padded. */ + if ((*ptr)->size < 8) + where += 8 - (*ptr)->size; +#endif + memcpy (where, *p_argv.c, (*ptr)->size); + next_arg.ul += words; + if (next_arg.ul == gpr_end.ul) + next_arg.ul = rest.ul; + } + } + break; + + case FFI_TYPE_UINT8: + gprvalue = **p_argv.uc; + goto putgpr; + case FFI_TYPE_SINT8: + gprvalue = **p_argv.sc; + goto putgpr; + case FFI_TYPE_UINT16: + gprvalue = **p_argv.us; + goto putgpr; + case FFI_TYPE_SINT16: + gprvalue = **p_argv.ss; + goto putgpr; + case FFI_TYPE_UINT32: + gprvalue = **p_argv.ui; + goto putgpr; + case FFI_TYPE_INT: + case FFI_TYPE_SINT32: + gprvalue = **p_argv.si; + goto putgpr; + + case FFI_TYPE_UINT64: + case FFI_TYPE_SINT64: + case FFI_TYPE_POINTER: + gprvalue = **p_argv.ul; + putgpr: + *next_arg.ul++ = gprvalue; + if (next_arg.ul == gpr_end.ul) + next_arg.ul = rest.ul; + break; + } + } + + FFI_ASSERT (flags & FLAG_4_GPR_ARGUMENTS + || (next_arg.ul >= gpr_base.ul + && next_arg.ul <= gpr_base.ul + 4)); +} + + +#if _CALL_ELF == 2 +#define MIN_CACHE_LINE_SIZE 8 + +static void +flush_icache (char *wraddr, char *xaddr, int size) +{ + int i; + for (i = 0; i < size; i += MIN_CACHE_LINE_SIZE) + __asm__ volatile ("icbi 0,%0;" "dcbf 0,%1;" + : : "r" (xaddr + i), "r" (wraddr + i) : "memory"); + __asm__ volatile ("icbi 0,%0;" "dcbf 0,%1;" "sync;" "isync;" + : : "r"(xaddr + size - 1), "r"(wraddr + size - 1) + : "memory"); +} +#endif + +ffi_status +ffi_prep_closure_loc_linux64 (ffi_closure *closure, + ffi_cif *cif, + void (*fun) (ffi_cif *, void *, void **, void *), + void *user_data, + void *codeloc) +{ +#if _CALL_ELF == 2 + unsigned int *tramp = (unsigned int *) &closure->tramp[0]; + + if (cif->abi < FFI_LINUX || cif->abi >= FFI_LAST_ABI) + return FFI_BAD_ABI; + + tramp[0] = 0xe96c0018; /* 0: ld 11,2f-0b(12) */ + tramp[1] = 0xe98c0010; /* ld 12,1f-0b(12) */ + tramp[2] = 0x7d8903a6; /* mtctr 12 */ + tramp[3] = 0x4e800420; /* bctr */ + /* 1: .quad function_addr */ + /* 2: .quad context */ + *(void **) &tramp[4] = (void *) ffi_closure_LINUX64; + *(void **) &tramp[6] = codeloc; + flush_icache ((char *)tramp, (char *)codeloc, FFI_TRAMPOLINE_SIZE); +#else + void **tramp = (void **) &closure->tramp[0]; + + if (cif->abi < FFI_LINUX || cif->abi >= FFI_LAST_ABI) + return FFI_BAD_ABI; + + /* Copy function address and TOC from ffi_closure_LINUX64. */ + memcpy (tramp, (char *) ffi_closure_LINUX64, 16); + tramp[2] = codeloc; +#endif + + closure->cif = cif; + closure->fun = fun; + closure->user_data = user_data; + + return FFI_OK; +} + + +int FFI_HIDDEN +ffi_closure_helper_LINUX64 (ffi_closure *closure, void *rvalue, + unsigned long *pst, ffi_dblfl *pfr) +{ + /* rvalue is the pointer to space for return value in closure assembly */ + /* pst is the pointer to parameter save area + (r3-r10 are stored into its first 8 slots by ffi_closure_LINUX64) */ + /* pfr is the pointer to where f1-f13 are stored in ffi_closure_LINUX64 */ + + void **avalue; + ffi_type **arg_types; + unsigned long i, avn, nfixedargs; + ffi_cif *cif; + ffi_dblfl *end_pfr = pfr + NUM_FPR_ARG_REGISTERS64; + unsigned long align; + + cif = closure->cif; + avalue = alloca (cif->nargs * sizeof (void *)); + + /* Copy the caller's structure return value address so that the + closure returns the data directly to the caller. */ + if (cif->rtype->type == FFI_TYPE_STRUCT + && (cif->flags & FLAG_RETURNS_SMST) == 0) + { + rvalue = (void *) *pst; + pst++; + } + + i = 0; + avn = cif->nargs; +#if _CALL_ELF != 2 + nfixedargs = (unsigned) -1; + if ((cif->flags & FLAG_COMPAT) == 0) +#endif + nfixedargs = cif->nfixedargs; + arg_types = cif->arg_types; + + /* Grab the addresses of the arguments from the stack frame. */ + while (i < avn) + { + unsigned int elt, elnum; + + switch (arg_types[i]->type) + { + case FFI_TYPE_SINT8: + case FFI_TYPE_UINT8: +#ifndef __LITTLE_ENDIAN__ + avalue[i] = (char *) pst + 7; + pst++; + break; +#endif + + case FFI_TYPE_SINT16: + case FFI_TYPE_UINT16: +#ifndef __LITTLE_ENDIAN__ + avalue[i] = (char *) pst + 6; + pst++; + break; +#endif + + case FFI_TYPE_SINT32: + case FFI_TYPE_UINT32: +#ifndef __LITTLE_ENDIAN__ + avalue[i] = (char *) pst + 4; + pst++; + break; +#endif + + case FFI_TYPE_SINT64: + case FFI_TYPE_UINT64: + case FFI_TYPE_POINTER: + avalue[i] = pst; + pst++; + break; + + case FFI_TYPE_STRUCT: + if ((cif->abi & FFI_LINUX_STRUCT_ALIGN) != 0) + { + align = arg_types[i]->alignment; + if (align > 16) + align = 16; + if (align > 1) + pst = (unsigned long *) ALIGN ((size_t) pst, align); + } + elt = 0; +#if _CALL_ELF == 2 + elt = discover_homogeneous_aggregate (arg_types[i], &elnum); +#endif + if (elt) + { + union { + void *v; + unsigned long *ul; + float *f; + double *d; + size_t p; + } to, from; + + /* Repackage the aggregate from its parts. The + aggregate size is not greater than the space taken by + the registers so store back to the register/parameter + save arrays. */ + if (pfr + elnum <= end_pfr) + to.v = pfr; + else + to.v = pst; + + avalue[i] = to.v; + from.ul = pst; + if (elt == FFI_TYPE_FLOAT) + { + do + { + if (pfr < end_pfr && i < nfixedargs) + { + *to.f = (float) pfr->d; + pfr++; + } + else + *to.f = *from.f; + to.f++; + from.f++; + } + while (--elnum != 0); + } + else + { + do + { + if (pfr < end_pfr && i < nfixedargs) + { + *to.d = pfr->d; + pfr++; + } + else + *to.d = *from.d; + to.d++; + from.d++; + } + while (--elnum != 0); + } + } + else + { +#ifndef __LITTLE_ENDIAN__ + /* Structures with size less than eight bytes are passed + left-padded. */ + if (arg_types[i]->size < 8) + avalue[i] = (char *) pst + 8 - arg_types[i]->size; + else +#endif + avalue[i] = pst; + } + pst += (arg_types[i]->size + 7) / 8; + break; + +#if FFI_TYPE_LONGDOUBLE != FFI_TYPE_DOUBLE + case FFI_TYPE_LONGDOUBLE: + if ((cif->abi & FFI_LINUX_LONG_DOUBLE_128) != 0) + { + if (pfr + 1 < end_pfr && i + 1 < nfixedargs) + { + avalue[i] = pfr; + pfr += 2; + } + else + { + if (pfr < end_pfr && i < nfixedargs) + { + /* Passed partly in f13 and partly on the stack. + Move it all to the stack. */ + *pst = *(unsigned long *) pfr; + pfr++; + } + avalue[i] = pst; + } + pst += 2; + break; + } + /* Fall through. */ +#endif + case FFI_TYPE_DOUBLE: + /* On the outgoing stack all values are aligned to 8 */ + /* there are 13 64bit floating point registers */ + + if (pfr < end_pfr && i < nfixedargs) + { + avalue[i] = pfr; + pfr++; + } + else + avalue[i] = pst; + pst++; + break; + + case FFI_TYPE_FLOAT: + if (pfr < end_pfr && i < nfixedargs) + { + /* Float values are stored as doubles in the + ffi_closure_LINUX64 code. Fix them here. */ + pfr->f = (float) pfr->d; + avalue[i] = pfr; + pfr++; + } + else + avalue[i] = pst; + pst++; + break; + + default: + FFI_ASSERT (0); + } + + i++; + } + + + (closure->fun) (cif, rvalue, avalue, closure->user_data); + + /* Tell ffi_closure_LINUX64 how to perform return type promotions. */ + if ((cif->flags & FLAG_RETURNS_SMST) != 0) + { + if ((cif->flags & FLAG_RETURNS_FP) == 0) + return FFI_V2_TYPE_SMALL_STRUCT + cif->rtype->size - 1; + else if ((cif->flags & FLAG_RETURNS_64BITS) != 0) + return FFI_V2_TYPE_DOUBLE_HOMOG; + else + return FFI_V2_TYPE_FLOAT_HOMOG; + } + return cif->rtype->type; +} +#endif Index: libffi-3.0.13/src/powerpc/ffi_powerpc.h =================================================================== --- /dev/null +++ libffi-3.0.13/src/powerpc/ffi_powerpc.h @@ -0,0 +1,77 @@ +/* ----------------------------------------------------------------------- + ffi_powerpc.h - Copyright (C) 2013 IBM + Copyright (C) 2011 Anthony Green + Copyright (C) 2011 Kyle Moffett + Copyright (C) 2008 Red Hat, Inc + Copyright (C) 2007, 2008 Free Software Foundation, Inc + Copyright (c) 1998 Geoffrey Keating + + PowerPC Foreign Function Interface + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + ``Software''), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice shall be included + in all copies or substantial portions of the Software. + + THE SOFTWARE IS PROVIDED ``AS IS'', WITHOUT WARRANTY OF ANY KIND, EXPRESS + OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY CLAIM, DAMAGES OR + OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + OTHER DEALINGS IN THE SOFTWARE. + ----------------------------------------------------------------------- */ + +enum { + /* The assembly depends on these exact flags. */ + /* These go in cr7 */ + FLAG_RETURNS_SMST = 1 << (31-31), /* Used for FFI_SYSV small structs. */ + FLAG_RETURNS_NOTHING = 1 << (31-30), + FLAG_RETURNS_FP = 1 << (31-29), + FLAG_RETURNS_64BITS = 1 << (31-28), + + /* This goes in cr6 */ + FLAG_RETURNS_128BITS = 1 << (31-27), + + FLAG_COMPAT = 1 << (31- 8), /* Not used by assembly */ + + /* These go in cr1 */ + FLAG_ARG_NEEDS_COPY = 1 << (31- 7), /* Used by sysv code */ + FLAG_ARG_NEEDS_PSAVE = FLAG_ARG_NEEDS_COPY, /* Used by linux64 code */ + FLAG_FP_ARGUMENTS = 1 << (31- 6), /* cr1.eq; specified by ABI */ + FLAG_4_GPR_ARGUMENTS = 1 << (31- 5), + FLAG_RETVAL_REFERENCE = 1 << (31- 4) +}; + +typedef union +{ + float f; + double d; +} ffi_dblfl; + +void FFI_HIDDEN ffi_closure_SYSV (void); +void FFI_HIDDEN ffi_call_SYSV(extended_cif *, unsigned, unsigned, unsigned *, + void (*)(void)); + +void FFI_HIDDEN ffi_prep_types_sysv (ffi_abi); +ffi_status FFI_HIDDEN ffi_prep_cif_sysv (ffi_cif *); +int FFI_HIDDEN ffi_closure_helper_SYSV (ffi_closure *, void *, unsigned long *, + ffi_dblfl *, unsigned long *); + +void FFI_HIDDEN ffi_call_LINUX64(extended_cif *, unsigned long, unsigned long, + unsigned long *, void (*)(void)); +void FFI_HIDDEN ffi_closure_LINUX64 (void); + +void FFI_HIDDEN ffi_prep_types_linux64 (ffi_abi); +ffi_status FFI_HIDDEN ffi_prep_cif_linux64 (ffi_cif *); +ffi_status FFI_HIDDEN ffi_prep_cif_linux64_var (ffi_cif *, unsigned int, + unsigned int); +void FFI_HIDDEN ffi_prep_args64 (extended_cif *, unsigned long *const); +int FFI_HIDDEN ffi_closure_helper_LINUX64 (ffi_closure *, void *, + unsigned long *, ffi_dblfl *); Index: libffi-3.0.13/src/powerpc/ffi_sysv.c =================================================================== --- /dev/null +++ libffi-3.0.13/src/powerpc/ffi_sysv.c @@ -0,0 +1,931 @@ +/* ----------------------------------------------------------------------- + ffi_sysv.c - Copyright (C) 2013 IBM + Copyright (C) 2011 Anthony Green + Copyright (C) 2011 Kyle Moffett + Copyright (C) 2008 Red Hat, Inc + Copyright (C) 2007, 2008 Free Software Foundation, Inc + Copyright (c) 1998 Geoffrey Keating + + PowerPC Foreign Function Interface + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + ``Software''), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice shall be included + in all copies or substantial portions of the Software. + + THE SOFTWARE IS PROVIDED ``AS IS'', WITHOUT WARRANTY OF ANY KIND, EXPRESS + OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY CLAIM, DAMAGES OR + OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + OTHER DEALINGS IN THE SOFTWARE. + ----------------------------------------------------------------------- */ + +#include "ffi.h" + +#ifndef POWERPC64 +#include "ffi_common.h" +#include "ffi_powerpc.h" + + +/* About the SYSV ABI. */ +#define ASM_NEEDS_REGISTERS 4 +#define NUM_GPR_ARG_REGISTERS 8 +#define NUM_FPR_ARG_REGISTERS 8 + + +#if HAVE_LONG_DOUBLE_VARIANT && FFI_TYPE_LONGDOUBLE != FFI_TYPE_DOUBLE +/* Adjust size of ffi_type_longdouble. */ +void FFI_HIDDEN +ffi_prep_types_sysv (ffi_abi abi) +{ + if ((abi & (FFI_SYSV | FFI_SYSV_LONG_DOUBLE_128)) == FFI_SYSV) + { + ffi_type_longdouble.size = 8; + ffi_type_longdouble.alignment = 8; + } + else + { + ffi_type_longdouble.size = 16; + ffi_type_longdouble.alignment = 16; + } +} +#endif + +/* Transform long double, double and float to other types as per abi. */ +static int +translate_float (int abi, int type) +{ +#if FFI_TYPE_LONGDOUBLE != FFI_TYPE_DOUBLE + if (type == FFI_TYPE_LONGDOUBLE + && (abi & FFI_SYSV_LONG_DOUBLE_128) == 0) + type = FFI_TYPE_DOUBLE; +#endif + if ((abi & FFI_SYSV_SOFT_FLOAT) != 0) + { + if (type == FFI_TYPE_FLOAT) + type = FFI_TYPE_UINT32; + else if (type == FFI_TYPE_DOUBLE) + type = FFI_TYPE_UINT64; +#if FFI_TYPE_LONGDOUBLE != FFI_TYPE_DOUBLE + else if (type == FFI_TYPE_LONGDOUBLE) + type = FFI_TYPE_UINT128; + } + else if ((abi & FFI_SYSV_IBM_LONG_DOUBLE) == 0) + { + if (type == FFI_TYPE_LONGDOUBLE) + type = FFI_TYPE_STRUCT; +#endif + } + return type; +} + +/* Perform machine dependent cif processing */ +static ffi_status +ffi_prep_cif_sysv_core (ffi_cif *cif) +{ + ffi_type **ptr; + unsigned bytes; + unsigned i, fparg_count = 0, intarg_count = 0; + unsigned flags = cif->flags; + unsigned struct_copy_size = 0; + unsigned type = cif->rtype->type; + unsigned size = cif->rtype->size; + + /* The machine-independent calculation of cif->bytes doesn't work + for us. Redo the calculation. */ + + /* Space for the frame pointer, callee's LR, and the asm's temp regs. */ + bytes = (2 + ASM_NEEDS_REGISTERS) * sizeof (int); + + /* Space for the GPR registers. */ + bytes += NUM_GPR_ARG_REGISTERS * sizeof (int); + + /* Return value handling. The rules for SYSV are as follows: + - 32-bit (or less) integer values are returned in gpr3; + - Structures of size <= 4 bytes also returned in gpr3; + - 64-bit integer values and structures between 5 and 8 bytes are returned + in gpr3 and gpr4; + - Larger structures are allocated space and a pointer is passed as + the first argument. + - Single/double FP values are returned in fpr1; + - long doubles (if not equivalent to double) are returned in + fpr1,fpr2 for Linux and as for large structs for SysV. */ + + type = translate_float (cif->abi, type); + + switch (type) + { +#if FFI_TYPE_LONGDOUBLE != FFI_TYPE_DOUBLE + case FFI_TYPE_LONGDOUBLE: + flags |= FLAG_RETURNS_128BITS; + /* Fall through. */ +#endif + case FFI_TYPE_DOUBLE: + flags |= FLAG_RETURNS_64BITS; + /* Fall through. */ + case FFI_TYPE_FLOAT: + flags |= FLAG_RETURNS_FP; +#ifdef __NO_FPRS__ + return FFI_BAD_ABI; +#endif + break; + + case FFI_TYPE_UINT128: + flags |= FLAG_RETURNS_128BITS; + /* Fall through. */ + case FFI_TYPE_UINT64: + case FFI_TYPE_SINT64: + flags |= FLAG_RETURNS_64BITS; + break; + + case FFI_TYPE_STRUCT: + /* The final SYSV ABI says that structures smaller or equal 8 bytes + are returned in r3/r4. A draft ABI used by linux instead + returns them in memory. */ + if ((cif->abi & FFI_SYSV_STRUCT_RET) != 0 && size <= 8) + { + flags |= FLAG_RETURNS_SMST; + break; + } + intarg_count++; + flags |= FLAG_RETVAL_REFERENCE; + /* Fall through. */ + case FFI_TYPE_VOID: + flags |= FLAG_RETURNS_NOTHING; + break; + + default: + /* Returns 32-bit integer, or similar. Nothing to do here. */ + break; + } + + /* The first NUM_GPR_ARG_REGISTERS words of integer arguments, and the + first NUM_FPR_ARG_REGISTERS fp arguments, go in registers; the rest + goes on the stack. Structures and long doubles (if not equivalent + to double) are passed as a pointer to a copy of the structure. + Stuff on the stack needs to keep proper alignment. */ + for (ptr = cif->arg_types, i = cif->nargs; i > 0; i--, ptr++) + { + unsigned short typenum = (*ptr)->type; + + typenum = translate_float (cif->abi, typenum); + + switch (typenum) + { +#if FFI_TYPE_LONGDOUBLE != FFI_TYPE_DOUBLE + case FFI_TYPE_LONGDOUBLE: + fparg_count++; + /* Fall thru */ +#endif + case FFI_TYPE_DOUBLE: + fparg_count++; + /* If this FP arg is going on the stack, it must be + 8-byte-aligned. */ + if (fparg_count > NUM_FPR_ARG_REGISTERS + && intarg_count >= NUM_GPR_ARG_REGISTERS + && intarg_count % 2 != 0) + intarg_count++; +#ifdef __NO_FPRS__ + return FFI_BAD_ABI; +#endif + break; + + case FFI_TYPE_FLOAT: + fparg_count++; +#ifdef __NO_FPRS__ + return FFI_BAD_ABI; +#endif + break; + + case FFI_TYPE_UINT128: + /* A long double in FFI_LINUX_SOFT_FLOAT can use only a set + of four consecutive gprs. If we do not have enough, we + have to adjust the intarg_count value. */ + if (intarg_count >= NUM_GPR_ARG_REGISTERS - 3 + && intarg_count < NUM_GPR_ARG_REGISTERS) + intarg_count = NUM_GPR_ARG_REGISTERS; + intarg_count += 4; + break; + + case FFI_TYPE_UINT64: + case FFI_TYPE_SINT64: + /* 'long long' arguments are passed as two words, but + either both words must fit in registers or both go + on the stack. If they go on the stack, they must + be 8-byte-aligned. + + Also, only certain register pairs can be used for + passing long long int -- specifically (r3,r4), (r5,r6), + (r7,r8), (r9,r10). */ + if (intarg_count == NUM_GPR_ARG_REGISTERS-1 + || intarg_count % 2 != 0) + intarg_count++; + intarg_count += 2; + break; + + case FFI_TYPE_STRUCT: + /* We must allocate space for a copy of these to enforce + pass-by-value. Pad the space up to a multiple of 16 + bytes (the maximum alignment required for anything under + the SYSV ABI). */ + struct_copy_size += ((*ptr)->size + 15) & ~0xF; + /* Fall through (allocate space for the pointer). */ + + case FFI_TYPE_POINTER: + case FFI_TYPE_INT: + case FFI_TYPE_UINT32: + case FFI_TYPE_SINT32: + case FFI_TYPE_UINT16: + case FFI_TYPE_SINT16: + case FFI_TYPE_UINT8: + case FFI_TYPE_SINT8: + /* Everything else is passed as a 4-byte word in a GPR, either + the object itself or a pointer to it. */ + intarg_count++; + break; + + default: + FFI_ASSERT (0); + } + } + + if (fparg_count != 0) + flags |= FLAG_FP_ARGUMENTS; + if (intarg_count > 4) + flags |= FLAG_4_GPR_ARGUMENTS; + if (struct_copy_size != 0) + flags |= FLAG_ARG_NEEDS_COPY; + + /* Space for the FPR registers, if needed. */ + if (fparg_count != 0) + bytes += NUM_FPR_ARG_REGISTERS * sizeof (double); + + /* Stack space. */ + if (intarg_count > NUM_GPR_ARG_REGISTERS) + bytes += (intarg_count - NUM_GPR_ARG_REGISTERS) * sizeof (int); + if (fparg_count > NUM_FPR_ARG_REGISTERS) + bytes += (fparg_count - NUM_FPR_ARG_REGISTERS) * sizeof (double); + + /* The stack space allocated needs to be a multiple of 16 bytes. */ + bytes = (bytes + 15) & ~0xF; + + /* Add in the space for the copied structures. */ + bytes += struct_copy_size; + + cif->flags = flags; + cif->bytes = bytes; + + return FFI_OK; +} + +ffi_status FFI_HIDDEN +ffi_prep_cif_sysv (ffi_cif *cif) +{ + if ((cif->abi & FFI_SYSV) == 0) + { + /* This call is from old code. Translate to new ABI values. */ + cif->flags |= FLAG_COMPAT; + switch (cif->abi) + { + default: + return FFI_BAD_ABI; + + case FFI_COMPAT_SYSV: + cif->abi = FFI_SYSV | FFI_SYSV_STRUCT_RET | FFI_SYSV_LONG_DOUBLE_128; + break; + + case FFI_COMPAT_GCC_SYSV: + cif->abi = FFI_SYSV | FFI_SYSV_LONG_DOUBLE_128; + break; + + case FFI_COMPAT_LINUX: + cif->abi = (FFI_SYSV | FFI_SYSV_IBM_LONG_DOUBLE + | FFI_SYSV_LONG_DOUBLE_128); + break; + + case FFI_COMPAT_LINUX_SOFT_FLOAT: + cif->abi = (FFI_SYSV | FFI_SYSV_SOFT_FLOAT | FFI_SYSV_IBM_LONG_DOUBLE + | FFI_SYSV_LONG_DOUBLE_128); + break; + } + } + return ffi_prep_cif_sysv_core (cif); +} + +/* ffi_prep_args_SYSV is called by the assembly routine once stack space + has been allocated for the function's arguments. + + The stack layout we want looks like this: + + | Return address from ffi_call_SYSV 4bytes | higher addresses + |--------------------------------------------| + | Previous backchain pointer 4 | stack pointer here + |--------------------------------------------|<+ <<< on entry to + | Saved r28-r31 4*4 | | ffi_call_SYSV + |--------------------------------------------| | + | GPR registers r3-r10 8*4 | | ffi_call_SYSV + |--------------------------------------------| | + | FPR registers f1-f8 (optional) 8*8 | | + |--------------------------------------------| | stack | + | Space for copied structures | | grows | + |--------------------------------------------| | down V + | Parameters that didn't fit in registers | | + |--------------------------------------------| | lower addresses + | Space for callee's LR 4 | | + |--------------------------------------------| | stack pointer here + | Current backchain pointer 4 |-/ during + |--------------------------------------------| <<< ffi_call_SYSV + +*/ + +void FFI_HIDDEN +ffi_prep_args_SYSV (extended_cif *ecif, unsigned *const stack) +{ + const unsigned bytes = ecif->cif->bytes; + const unsigned flags = ecif->cif->flags; + + typedef union + { + char *c; + unsigned *u; + long long *ll; + float *f; + double *d; + } valp; + + /* 'stacktop' points at the previous backchain pointer. */ + valp stacktop; + + /* 'gpr_base' points at the space for gpr3, and grows upwards as + we use GPR registers. */ + valp gpr_base; + int intarg_count; + +#ifndef __NO_FPRS__ + /* 'fpr_base' points at the space for fpr1, and grows upwards as + we use FPR registers. */ + valp fpr_base; + int fparg_count; +#endif + + /* 'copy_space' grows down as we put structures in it. It should + stay 16-byte aligned. */ + valp copy_space; + + /* 'next_arg' grows up as we put parameters in it. */ + valp next_arg; + + int i; + ffi_type **ptr; +#ifndef __NO_FPRS__ + double double_tmp; +#endif + union + { + void **v; + char **c; + signed char **sc; + unsigned char **uc; + signed short **ss; + unsigned short **us; + unsigned int **ui; + long long **ll; + float **f; + double **d; + } p_argv; + size_t struct_copy_size; + unsigned gprvalue; + + stacktop.c = (char *) stack + bytes; + gpr_base.u = stacktop.u - ASM_NEEDS_REGISTERS - NUM_GPR_ARG_REGISTERS; + intarg_count = 0; +#ifndef __NO_FPRS__ + fpr_base.d = gpr_base.d - NUM_FPR_ARG_REGISTERS; + fparg_count = 0; + copy_space.c = ((flags & FLAG_FP_ARGUMENTS) ? fpr_base.c : gpr_base.c); +#else + copy_space.c = gpr_base.c; +#endif + next_arg.u = stack + 2; + + /* Check that everything starts aligned properly. */ + FFI_ASSERT (((unsigned long) (char *) stack & 0xF) == 0); + FFI_ASSERT (((unsigned long) copy_space.c & 0xF) == 0); + FFI_ASSERT (((unsigned long) stacktop.c & 0xF) == 0); + FFI_ASSERT ((bytes & 0xF) == 0); + FFI_ASSERT (copy_space.c >= next_arg.c); + + /* Deal with return values that are actually pass-by-reference. */ + if (flags & FLAG_RETVAL_REFERENCE) + { + *gpr_base.u++ = (unsigned long) (char *) ecif->rvalue; + intarg_count++; + } + + /* Now for the arguments. */ + p_argv.v = ecif->avalue; + for (ptr = ecif->cif->arg_types, i = ecif->cif->nargs; + i > 0; + i--, ptr++, p_argv.v++) + { + unsigned int typenum = (*ptr)->type; + + typenum = translate_float (ecif->cif->abi, typenum); + + /* Now test the translated value */ + switch (typenum) + { +#ifndef __NO_FPRS__ +# if FFI_TYPE_LONGDOUBLE != FFI_TYPE_DOUBLE + case FFI_TYPE_LONGDOUBLE: + double_tmp = (*p_argv.d)[0]; + + if (fparg_count >= NUM_FPR_ARG_REGISTERS - 1) + { + if (intarg_count >= NUM_GPR_ARG_REGISTERS + && intarg_count % 2 != 0) + { + intarg_count++; + next_arg.u++; + } + *next_arg.d = double_tmp; + next_arg.u += 2; + double_tmp = (*p_argv.d)[1]; + *next_arg.d = double_tmp; + next_arg.u += 2; + } + else + { + *fpr_base.d++ = double_tmp; + double_tmp = (*p_argv.d)[1]; + *fpr_base.d++ = double_tmp; + } + + fparg_count += 2; + FFI_ASSERT (flags & FLAG_FP_ARGUMENTS); + break; +# endif + case FFI_TYPE_DOUBLE: + double_tmp = **p_argv.d; + + if (fparg_count >= NUM_FPR_ARG_REGISTERS) + { + if (intarg_count >= NUM_GPR_ARG_REGISTERS + && intarg_count % 2 != 0) + { + intarg_count++; + next_arg.u++; + } + *next_arg.d = double_tmp; + next_arg.u += 2; + } + else + *fpr_base.d++ = double_tmp; + fparg_count++; + FFI_ASSERT (flags & FLAG_FP_ARGUMENTS); + break; + + case FFI_TYPE_FLOAT: + double_tmp = **p_argv.f; + if (fparg_count >= NUM_FPR_ARG_REGISTERS) + { + *next_arg.f = (float) double_tmp; + next_arg.u += 1; + intarg_count++; + } + else + *fpr_base.d++ = double_tmp; + fparg_count++; + FFI_ASSERT (flags & FLAG_FP_ARGUMENTS); + break; +#endif /* have FPRs */ + + case FFI_TYPE_UINT128: + /* The soft float ABI for long doubles works like this, a long double + is passed in four consecutive GPRs if available. A maximum of 2 + long doubles can be passed in gprs. If we do not have 4 GPRs + left, the long double is passed on the stack, 4-byte aligned. */ + { + unsigned int int_tmp; + unsigned int ii; + if (intarg_count >= NUM_GPR_ARG_REGISTERS - 3) + { + if (intarg_count < NUM_GPR_ARG_REGISTERS) + intarg_count = NUM_GPR_ARG_REGISTERS; + for (ii = 0; ii < 4; ii++) + { + int_tmp = (*p_argv.ui)[ii]; + *next_arg.u++ = int_tmp; + } + } + else + { + for (ii = 0; ii < 4; ii++) + { + int_tmp = (*p_argv.ui)[ii]; + *gpr_base.u++ = int_tmp; + } + } + intarg_count += 4; + break; + } + + case FFI_TYPE_UINT64: + case FFI_TYPE_SINT64: + if (intarg_count == NUM_GPR_ARG_REGISTERS-1) + intarg_count++; + if (intarg_count >= NUM_GPR_ARG_REGISTERS) + { + if (intarg_count % 2 != 0) + { + intarg_count++; + next_arg.u++; + } + *next_arg.ll = **p_argv.ll; + next_arg.u += 2; + } + else + { + /* The abi states only certain register pairs can be + used for passing long long int specifically (r3,r4), + (r5,r6), (r7,r8), (r9,r10). If next arg is long long + but not correct starting register of pair then skip + until the proper starting register. */ + if (intarg_count % 2 != 0) + { + intarg_count ++; + gpr_base.u++; + } + *gpr_base.ll++ = **p_argv.ll; + } + intarg_count += 2; + break; + + case FFI_TYPE_STRUCT: + struct_copy_size = ((*ptr)->size + 15) & ~0xF; + copy_space.c -= struct_copy_size; + memcpy (copy_space.c, *p_argv.c, (*ptr)->size); + + gprvalue = (unsigned long) copy_space.c; + + FFI_ASSERT (copy_space.c > next_arg.c); + FFI_ASSERT (flags & FLAG_ARG_NEEDS_COPY); + goto putgpr; + + case FFI_TYPE_UINT8: + gprvalue = **p_argv.uc; + goto putgpr; + case FFI_TYPE_SINT8: + gprvalue = **p_argv.sc; + goto putgpr; + case FFI_TYPE_UINT16: + gprvalue = **p_argv.us; + goto putgpr; + case FFI_TYPE_SINT16: + gprvalue = **p_argv.ss; + goto putgpr; + + case FFI_TYPE_INT: + case FFI_TYPE_UINT32: + case FFI_TYPE_SINT32: + case FFI_TYPE_POINTER: + + gprvalue = **p_argv.ui; + + putgpr: + if (intarg_count >= NUM_GPR_ARG_REGISTERS) + *next_arg.u++ = gprvalue; + else + *gpr_base.u++ = gprvalue; + intarg_count++; + break; + } + } + + /* Check that we didn't overrun the stack... */ + FFI_ASSERT (copy_space.c >= next_arg.c); + FFI_ASSERT (gpr_base.u <= stacktop.u - ASM_NEEDS_REGISTERS); + /* The assert below is testing that the number of integer arguments agrees + with the number found in ffi_prep_cif_machdep(). However, intarg_count + is incremented whenever we place an FP arg on the stack, so account for + that before our assert test. */ +#ifndef __NO_FPRS__ + if (fparg_count > NUM_FPR_ARG_REGISTERS) + intarg_count -= fparg_count - NUM_FPR_ARG_REGISTERS; + FFI_ASSERT (fpr_base.u + <= stacktop.u - ASM_NEEDS_REGISTERS - NUM_GPR_ARG_REGISTERS); +#endif + FFI_ASSERT (flags & FLAG_4_GPR_ARGUMENTS || intarg_count <= 4); +} + +#define MIN_CACHE_LINE_SIZE 8 + +static void +flush_icache (char *wraddr, char *xaddr, int size) +{ + int i; + for (i = 0; i < size; i += MIN_CACHE_LINE_SIZE) + __asm__ volatile ("icbi 0,%0;" "dcbf 0,%1;" + : : "r" (xaddr + i), "r" (wraddr + i) : "memory"); + __asm__ volatile ("icbi 0,%0;" "dcbf 0,%1;" "sync;" "isync;" + : : "r"(xaddr + size - 1), "r"(wraddr + size - 1) + : "memory"); +} + +ffi_status FFI_HIDDEN +ffi_prep_closure_loc_sysv (ffi_closure *closure, + ffi_cif *cif, + void (*fun) (ffi_cif *, void *, void **, void *), + void *user_data, + void *codeloc) +{ + unsigned int *tramp; + + if (cif->abi < FFI_SYSV || cif->abi >= FFI_LAST_ABI) + return FFI_BAD_ABI; + + tramp = (unsigned int *) &closure->tramp[0]; + tramp[0] = 0x7c0802a6; /* mflr r0 */ + tramp[1] = 0x4800000d; /* bl 10 */ + tramp[4] = 0x7d6802a6; /* mflr r11 */ + tramp[5] = 0x7c0803a6; /* mtlr r0 */ + tramp[6] = 0x800b0000; /* lwz r0,0(r11) */ + tramp[7] = 0x816b0004; /* lwz r11,4(r11) */ + tramp[8] = 0x7c0903a6; /* mtctr r0 */ + tramp[9] = 0x4e800420; /* bctr */ + *(void **) &tramp[2] = (void *) ffi_closure_SYSV; /* function */ + *(void **) &tramp[3] = codeloc; /* context */ + + /* Flush the icache. */ + flush_icache ((char *)tramp, (char *)codeloc, FFI_TRAMPOLINE_SIZE); + + closure->cif = cif; + closure->fun = fun; + closure->user_data = user_data; + + return FFI_OK; +} + +/* Basically the trampoline invokes ffi_closure_SYSV, and on + entry, r11 holds the address of the closure. + After storing the registers that could possibly contain + parameters to be passed into the stack frame and setting + up space for a return value, ffi_closure_SYSV invokes the + following helper function to do most of the work. */ + +int +ffi_closure_helper_SYSV (ffi_closure *closure, void *rvalue, + unsigned long *pgr, ffi_dblfl *pfr, + unsigned long *pst) +{ + /* rvalue is the pointer to space for return value in closure assembly */ + /* pgr is the pointer to where r3-r10 are stored in ffi_closure_SYSV */ + /* pfr is the pointer to where f1-f8 are stored in ffi_closure_SYSV */ + /* pst is the pointer to outgoing parameter stack in original caller */ + + void ** avalue; + ffi_type ** arg_types; + long i, avn; +#ifndef __NO_FPRS__ + long nf = 0; /* number of floating registers already used */ +#endif + long ng = 0; /* number of general registers already used */ + + ffi_cif *cif = closure->cif; + unsigned size = cif->rtype->size; + unsigned short rtypenum = cif->rtype->type; + + avalue = alloca (cif->nargs * sizeof (void *)); + + /* First translate for softfloat/nonlinux */ + rtypenum = translate_float (cif->abi, rtypenum); + + /* Copy the caller's structure return value address so that the closure + returns the data directly to the caller. + For FFI_SYSV the result is passed in r3/r4 if the struct size is less + or equal 8 bytes. */ + if (rtypenum == FFI_TYPE_STRUCT + && !((cif->abi & FFI_SYSV_STRUCT_RET) != 0 && size <= 8)) + { + rvalue = (void *) *pgr; + ng++; + pgr++; + } + + i = 0; + avn = cif->nargs; + arg_types = cif->arg_types; + + /* Grab the addresses of the arguments from the stack frame. */ + while (i < avn) { + unsigned short typenum = arg_types[i]->type; + + /* We may need to handle some values depending on ABI. */ + typenum = translate_float (cif->abi, typenum); + + switch (typenum) + { +#ifndef __NO_FPRS__ + case FFI_TYPE_FLOAT: + /* Unfortunately float values are stored as doubles + in the ffi_closure_SYSV code (since we don't check + the type in that routine). */ + if (nf < NUM_FPR_ARG_REGISTERS) + { + /* FIXME? here we are really changing the values + stored in the original calling routines outgoing + parameter stack. This is probably a really + naughty thing to do but... */ + double temp = pfr->d; + pfr->f = (float) temp; + avalue[i] = pfr; + nf++; + pfr++; + } + else + { + avalue[i] = pst; + pst += 1; + } + break; + + case FFI_TYPE_DOUBLE: + if (nf < NUM_FPR_ARG_REGISTERS) + { + avalue[i] = pfr; + nf++; + pfr++; + } + else + { + if (((long) pst) & 4) + pst++; + avalue[i] = pst; + pst += 2; + } + break; + +# if FFI_TYPE_LONGDOUBLE != FFI_TYPE_DOUBLE + case FFI_TYPE_LONGDOUBLE: + if (nf < NUM_FPR_ARG_REGISTERS - 1) + { + avalue[i] = pfr; + pfr += 2; + nf += 2; + } + else + { + if (((long) pst) & 4) + pst++; + avalue[i] = pst; + pst += 4; + nf = 8; + } + break; +# endif +#endif + + case FFI_TYPE_UINT128: + /* Test if for the whole long double, 4 gprs are available. + otherwise the stuff ends up on the stack. */ + if (ng < NUM_GPR_ARG_REGISTERS - 3) + { + avalue[i] = pgr; + pgr += 4; + ng += 4; + } + else + { + avalue[i] = pst; + pst += 4; + ng = 8+4; + } + break; + + case FFI_TYPE_SINT8: + case FFI_TYPE_UINT8: +#ifndef __LITTLE_ENDIAN__ + if (ng < NUM_GPR_ARG_REGISTERS) + { + avalue[i] = (char *) pgr + 3; + ng++; + pgr++; + } + else + { + avalue[i] = (char *) pst + 3; + pst++; + } + break; +#endif + + case FFI_TYPE_SINT16: + case FFI_TYPE_UINT16: +#ifndef __LITTLE_ENDIAN__ + if (ng < NUM_GPR_ARG_REGISTERS) + { + avalue[i] = (char *) pgr + 2; + ng++; + pgr++; + } + else + { + avalue[i] = (char *) pst + 2; + pst++; + } + break; +#endif + + case FFI_TYPE_SINT32: + case FFI_TYPE_UINT32: + case FFI_TYPE_POINTER: + if (ng < NUM_GPR_ARG_REGISTERS) + { + avalue[i] = pgr; + ng++; + pgr++; + } + else + { + avalue[i] = pst; + pst++; + } + break; + + case FFI_TYPE_STRUCT: + /* Structs are passed by reference. The address will appear in a + gpr if it is one of the first 8 arguments. */ + if (ng < NUM_GPR_ARG_REGISTERS) + { + avalue[i] = (void *) *pgr; + ng++; + pgr++; + } + else + { + avalue[i] = (void *) *pst; + pst++; + } + break; + + case FFI_TYPE_SINT64: + case FFI_TYPE_UINT64: + /* Passing long long ints are complex, they must + be passed in suitable register pairs such as + (r3,r4) or (r5,r6) or (r6,r7), or (r7,r8) or (r9,r10) + and if the entire pair aren't available then the outgoing + parameter stack is used for both but an alignment of 8 + must will be kept. So we must either look in pgr + or pst to find the correct address for this type + of parameter. */ + if (ng < NUM_GPR_ARG_REGISTERS - 1) + { + if (ng & 1) + { + /* skip r4, r6, r8 as starting points */ + ng++; + pgr++; + } + avalue[i] = pgr; + ng += 2; + pgr += 2; + } + else + { + if (((long) pst) & 4) + pst++; + avalue[i] = pst; + pst += 2; + ng = NUM_GPR_ARG_REGISTERS; + } + break; + + default: + FFI_ASSERT (0); + } + + i++; + } + + (closure->fun) (cif, rvalue, avalue, closure->user_data); + + /* Tell ffi_closure_SYSV how to perform return type promotions. + Because the FFI_SYSV ABI returns the structures <= 8 bytes in + r3/r4 we have to tell ffi_closure_SYSV how to treat them. We + combine the base type FFI_SYSV_TYPE_SMALL_STRUCT with the size of + the struct less one. We never have a struct with size zero. + See the comment in ffitarget.h about ordering. */ + if (rtypenum == FFI_TYPE_STRUCT + && (cif->abi & FFI_SYSV_STRUCT_RET) != 0 && size <= 8) + return FFI_SYSV_TYPE_SMALL_STRUCT - 1 + size; + return rtypenum; +} +#endif Index: libffi-3.0.13/src/powerpc/ffitarget.h =================================================================== --- libffi-3.0.13.orig/src/powerpc/ffitarget.h +++ libffi-3.0.13/src/powerpc/ffitarget.h @@ -60,45 +60,76 @@ typedef signed long ffi_sarg; typedef enum ffi_abi { FFI_FIRST_ABI = 0, -#ifdef POWERPC - FFI_SYSV, - FFI_GCC_SYSV, - FFI_LINUX64, - FFI_LINUX, - FFI_LINUX_SOFT_FLOAT, -# if defined(POWERPC64) - FFI_DEFAULT_ABI = FFI_LINUX64, -# elif defined(__NO_FPRS__) - FFI_DEFAULT_ABI = FFI_LINUX_SOFT_FLOAT, -# elif (__LDBL_MANT_DIG__ == 106) - FFI_DEFAULT_ABI = FFI_LINUX, -# else - FFI_DEFAULT_ABI = FFI_GCC_SYSV, -# endif -#endif - -#ifdef POWERPC_AIX +#if defined (POWERPC_AIX) FFI_AIX, FFI_DARWIN, FFI_DEFAULT_ABI = FFI_AIX, -#endif + FFI_LAST_ABI -#ifdef POWERPC_DARWIN +#elif defined (POWERPC_DARWIN) FFI_AIX, FFI_DARWIN, FFI_DEFAULT_ABI = FFI_DARWIN, -#endif + FFI_LAST_ABI + +#else + /* The FFI_COMPAT values are used by old code. Since libffi may be + a shared library we have to support old values for backwards + compatibility. */ + FFI_COMPAT_SYSV, + FFI_COMPAT_GCC_SYSV, + FFI_COMPAT_LINUX64, + FFI_COMPAT_LINUX, + FFI_COMPAT_LINUX_SOFT_FLOAT, + +# if defined (POWERPC64) + /* This bit, always set in new code, must not be set in any of the + old FFI_COMPAT values that might be used for 64-bit linux. We + only need worry about FFI_COMPAT_LINUX64, but to be safe avoid + all old values. */ + FFI_LINUX = 8, + /* This and following bits can reuse FFI_COMPAT values. */ + FFI_LINUX_STRUCT_ALIGN = 1, + FFI_LINUX_LONG_DOUBLE_128 = 2, + FFI_DEFAULT_ABI = (FFI_LINUX +# ifdef __STRUCT_PARM_ALIGN__ + | FFI_LINUX_STRUCT_ALIGN +# endif +# ifdef __LONG_DOUBLE_128__ + | FFI_LINUX_LONG_DOUBLE_128 +# endif + ), + FFI_LAST_ABI = 12 -#ifdef POWERPC_FREEBSD - FFI_SYSV, - FFI_GCC_SYSV, - FFI_LINUX64, - FFI_LINUX, - FFI_LINUX_SOFT_FLOAT, - FFI_DEFAULT_ABI = FFI_SYSV, +# else + /* This bit, always set in new code, must not be set in any of the + old FFI_COMPAT values that might be used for 32-bit linux/sysv/bsd. */ + FFI_SYSV = 8, + /* This and following bits can reuse FFI_COMPAT values. */ + FFI_SYSV_SOFT_FLOAT = 1, + FFI_SYSV_STRUCT_RET = 2, + FFI_SYSV_IBM_LONG_DOUBLE = 4, + FFI_SYSV_LONG_DOUBLE_128 = 16, + + FFI_DEFAULT_ABI = (FFI_SYSV +# ifdef __NO_FPRS__ + | FFI_SYSV_SOFT_FLOAT +# endif +# if (defined (__SVR4_STRUCT_RETURN) \ + || defined (POWERPC_FREEBSD) && !defined (__AIX_STRUCT_RETURN)) + | FFI_SYSV_STRUCT_RET +# endif +# if __LDBL_MANT_DIG__ == 106 + | FFI_SYSV_IBM_LONG_DOUBLE +# endif +# ifdef __LONG_DOUBLE_128__ + | FFI_SYSV_LONG_DOUBLE_128 +# endif + ), + FFI_LAST_ABI = 32 +# endif #endif - FFI_LAST_ABI } ffi_abi; #endif @@ -117,9 +148,7 @@ typedef enum ffi_abi { /* Needed for soft-float long-double-128 support. */ #define FFI_TYPE_UINT128 (FFI_TYPE_LAST + 1) -/* Needed for FFI_SYSV small structure returns. - We use two flag bits, (FLAG_SYSV_SMST_R3, FLAG_SYSV_SMST_R4) which are - defined in ffi.c, to determine the exact return type and its size. */ +/* Needed for FFI_SYSV small structure returns. */ #define FFI_SYSV_TYPE_SMALL_STRUCT (FFI_TYPE_LAST + 2) /* Used by ELFv2 for homogenous structure returns. */ Index: libffi-3.0.13/src/powerpc/linux64.S =================================================================== --- libffi-3.0.13.orig/src/powerpc/linux64.S +++ libffi-3.0.13/src/powerpc/linux64.S @@ -29,7 +29,7 @@ #include #include -#ifdef __powerpc64__ +#ifdef POWERPC64 .hidden ffi_call_LINUX64 .globl ffi_call_LINUX64 # if _CALL_ELF == 2 Index: libffi-3.0.13/src/powerpc/linux64_closure.S =================================================================== --- libffi-3.0.13.orig/src/powerpc/linux64_closure.S +++ libffi-3.0.13/src/powerpc/linux64_closure.S @@ -30,7 +30,7 @@ .file "linux64_closure.S" -#ifdef __powerpc64__ +#ifdef POWERPC64 FFI_HIDDEN (ffi_closure_LINUX64) .globl ffi_closure_LINUX64 # if _CALL_ELF == 2 Index: libffi-3.0.13/src/powerpc/ppc_closure.S =================================================================== --- libffi-3.0.13.orig/src/powerpc/ppc_closure.S +++ libffi-3.0.13/src/powerpc/ppc_closure.S @@ -31,7 +31,7 @@ .file "ppc_closure.S" -#ifndef __powerpc64__ +#ifndef POWERPC64 ENTRY(ffi_closure_SYSV) .LFB1: @@ -378,8 +378,7 @@ END(ffi_closure_SYSV) .align 2 .LEFDE1: -#endif - #if defined __ELF__ && defined __linux__ .section .note.GNU-stack,"",@progbits #endif +#endif Index: libffi-3.0.13/src/powerpc/sysv.S =================================================================== --- libffi-3.0.13.orig/src/powerpc/sysv.S +++ libffi-3.0.13/src/powerpc/sysv.S @@ -30,7 +30,7 @@ #include #include -#ifndef __powerpc64__ +#ifndef POWERPC64 .globl ffi_prep_args_SYSV ENTRY(ffi_call_SYSV) .LFB1: @@ -213,8 +213,8 @@ END(ffi_call_SYSV) .uleb128 0x1c .align 2 .LEFDE1: -#endif #if defined __ELF__ && defined __linux__ .section .note.GNU-stack,"",@progbits #endif +#endif Index: libffi-3.0.13/src/prep_cif.c =================================================================== --- libffi-3.0.13.orig/src/prep_cif.c +++ libffi-3.0.13/src/prep_cif.c @@ -126,6 +126,10 @@ ffi_status FFI_HIDDEN ffi_prep_cif_core( cif->flags = 0; +#if HAVE_LONG_DOUBLE_VARIANT + ffi_prep_types (abi); +#endif + /* Initialize the return type if necessary */ if ((cif->rtype->size == 0) && (initialize_aggregate(cif->rtype) != FFI_OK)) return FFI_BAD_TYPEDEF; Index: libffi-3.0.13/src/types.c =================================================================== --- libffi-3.0.13.orig/src/types.c +++ libffi-3.0.13/src/types.c @@ -44,6 +44,17 @@ const ffi_type ffi_type_##name = { \ id, NULL \ } +#define FFI_NONCONST_TYPEDEF(name, type, id) \ +struct struct_align_##name { \ + char c; \ + type x; \ +}; \ +ffi_type ffi_type_##name = { \ + sizeof(type), \ + offsetof(struct struct_align_##name, x), \ + id, NULL \ +} + /* Size and alignment are fake here. They must not be 0. */ const ffi_type ffi_type_void = { 1, 1, FFI_TYPE_VOID, NULL @@ -73,5 +84,9 @@ FFI_TYPEDEF(double, double, FFI_TYPE_DOU # endif const ffi_type ffi_type_longdouble = { 16, 16, 4, NULL }; #elif FFI_TYPE_LONGDOUBLE != FFI_TYPE_DOUBLE +# if HAVE_LONG_DOUBLE_VARIANT +FFI_NONCONST_TYPEDEF(longdouble, long double, FFI_TYPE_LONGDOUBLE); +# else FFI_TYPEDEF(longdouble, long double, FFI_TYPE_LONGDOUBLE); +# endif #endif